[Deck] Goblins

[magicmerl]

Maybe one more tin street hooligan could help against Junk ( not Jund ). But the problem is that equipments are only one of the problems we face against that matchup : at least the list i faced yesterday played tarmogoyf, knight of the reliquary, abrupt decay, maelstrom pulse, vindicate, liliana,hymn to tourach, thoughtseize.....with a good chance of playing his three drops turn two thanks to deathrite shaman and mox diamond. I suppose we must only hope that they get mana screwed to have a chance of winning against all the big critters, removals and disruptions.

That said, an active jitte is always the most problematic card for us, so additional artifact hate may be ok.

I'm not sure I agree with that logic. Because of the presence of Goblin Matron we effectively have 5 copies of TSH. Going from 5 copies up to 6 copies is a pretty marginal increase that comes at the expense of a precious SB slot. Plus we haev Sharpshooter and Prospector which also protect us from being hit by big creatures/equipment.

Of the cards you list that are a problem, graveyard hate nerfs DRS, Goyf and KotR. Thoughtseize, Abrupt Decay, Maelstron Pulse and Vindicate are just 1-for-1 answers that we don't care about because we have Ringleader/SGC. Liliana is bad because we have haste. Hymn is a little annoying but only one card.

So from my perspective I'd go with graveyard hate to strategically nerf their win conditions, and play out the games as if they are about grindy attrition.

[GoboLord]

Hey there fellow Warchiefs, Goblin-enthousiasts and lovers of statistics!
The present article is about a statistical approach to find out what makes Vial-Goblin decklists successful in legacy. Before we dive into the details, I’ll give you a brief overview of the structure of the present article.
1. theory - first I want to give you an idea of what exactly I want to find out (i.e. what kind of questions I want to answer with the data)
2. data-collection – second I briefly explain how I collected the data. This should give you a clear picture of what the data CAN tell us and what it CAN’T tell us
3. results & interpretation – given the massive output of results I present a summary of the most meaningful results, as well as my take on their interpretation
4. afterword

I. theory
Simply speaking, the purpose of the analyses is to find out what makes goblin decklists successful in legacy. To be more precise: I want to know which main-deck cards and card-configurations are indicators of success, irrespective of time, metagame, matchup and individual skill. Let me give you an example. Let’s assume I wanted to know how the number of Mogg War Marshal in the maindeck correlates with success. What I do then, is compare the average win-% of lists with 0 vs. 1 vs. 2 vs. 3 vs. 4 copies of MWM. This is, of course, a very global question, as we will figure out later. A more specific (“local”) question would be: “how does the number of MWMs correlate with success in mono-colored lists compared to multicolored lists?” Here are some of the questions I approached:
1. How does the number of Lands_Total/Basiclands/Fetchlands/Duals/Rishadan Ports/MWM/Warren Instigator/Piledriver/Warchief/Chieftain/Sharpshooter/Skirk Prospector/Tuktuk Scrapper/Tin Street Hooligan/Siege-Gang Commander/Krenko/Thalia/Gempalm Incinerator/Lightning Bolt/Tarfire/Stingscourger affect performance of the decklist?
2. How does the number of CMC=2 spells affect performance?
3. Does the data confirm the typology that I suggested in my last article (WInstigator vs. Classic vs. Thalia vs. Spaghetti)? And if not, what categories does the data reveal?
4. How does the number of “things to do by turn 2” affect performance?

II. data-collection
What kind of data did I collect? Well, on the basis of tournament reports that I could find on the internet I recorded the success of each goblin decklist I could find. “success” was measured as the outcome of a best-of-3 match and it tells you what % of the games were won. Example: decklist X lost round 1 of a tournament with a straight 0:2. This means that the success rate is 0%. Now the same deck wins round 2 with 2:1. This creatures a new record with 66,66% success rate (3 matches divided by 2 wins). So, each MATCH makes one record. By doing so the analysis weights the importance of larger tournaments, compared to smaller tournaments. Example: the results of a decklist that was played in 8 rounds of swiss create 8 records , whereas a decklist that was played on a small LGS-tourney with only 3 rounds provides 3 records.
I did not include the following cards in the analysis because I wasn’t able to find enough lists that featured LESS THAN 4 COPIES of them:
Aether Vial, Goblin Lackey, Goblin Warchief, Goblin Matron, Goblin Ringleader, Cavern of Souls, Wasteland
In the same sense I did not record sideboard cards, because their huge variety would make it difficult to collect a sufficient amount of data to get meaningful results. (You can accuse me for laziness if you want to :-D).

Below you can find a list of definitions of terms that I might use throughout the rest of the article.

* match: a best-of-3 match
* game: 2-3 games make one match
* success-rate or perfromance: win-%, percentage of games that were won
* N : “number”, like in “N=63” means that 63 decklists were running a certain configuration of cards.
* 2-drops: Goblin Piledriver, Mogg War Marshal, Warren Instigator, Thalia
* Removal: Stingscourger, Gempalm Incinerator, Lightning Bolt, Tarfire
* things to do by turn 2: any card that can be used by turn 2 without the help of Goblin lackey, Warchief, or Chrome Mox, which is: Goblin Lackey, Aether Vial, Skirk Prospector, Lightning Bolt, Tarfire, Gempalm Incinerator, Stingscourger, Thalia, Mogg War Marshal, Warren Instigator, Wasteland, Rishadan Port

III. Summary of results:

Let’s first look at how single cards affect performance, irrespective of the type of Goblin list you are playing. First come the results of the whole data-set, with which we have to compare the results of all single-cards.

TOTAL mean win-%: 62.77%
TOTAL matches recorded: 409
TOTAL number of decklists: 73

Rishadan Port - Lists with either 2 (+3,9%) or 4 (+0,5%) Ports performed above average.
Dual lands – Lists with either 1 (+3,38%) or 3 (+4,49 %) Duallands performed above average. MonoR lists performed slightly below average (-0,53%)
Goblin Piledriver – Lists with 0 (+0,12%), 2 (+1,5%) or 4 (+1,05%) Piledrivers performed above average. Lists with 3 Piledrivers performed slightly below average (-0,17%).
Mogg War Marshal – Lists with 3 (+0,23%) or 4 (+2,52%) MWMs performed above average.
Stingscourger – Lists with 1 Stingscourger performed above average (+0,53%). Lists with more or less than 1 Stingscourger performed below average.
Gempalm Incinerator – Lists with 4 GIs performed above average (+0,41%). Lists with less than 4 copies performed below average (-0,24% to -1,05%).
Goblin Chieftain – Lists with either 0 (+2,16%) , 1 (+3,78) or 4 (+0,95%) Chieftains performed above average.
Sharpshooter – Lists with 1 Skirk Prospector (+0,86%) performed slightly better than those with 0 copies (-1,23%).
Skirk Prospector - Lists with 1 Sharpshooter (+0,4%) performed slightly better than those with 0 copies (-0,55%)
Tin Street Hooligan or Tuktuk Scrapper – Lists with 1 “Shatter”-Goblin performed better (+2,06%) than those with 0 copies (-2,7%).
Thalia, Guardian of Thraben – Lists with 0 (+0,12%) Thalias performed better than those with 3 copies (-2,27%).
Krenko, Mob Boss and Siege-Gang Commander
0/2-split: -0,82% below avg
1/0-split: -1,52% below avg
1/1-split: -0,9% below avg
1/2-split: +3,14% above avg
There was a insufficient number of cases for all other possible splits.

Interpretation of results
* I can’t tell what exactly makes the list with 2 Rishadan Ports more successful than the rest. Further exploration revealed that about half of those lists were MonoR (62,90% wins) and other half was multi-colored (67,90% wins), while only very few list had either Tin Street Hooligan and/or Thalia in MD. My conclusion would that the 2-Port-lists were cutting Ports in favor of their splashes, which they used for something that I didn’t cover with my analysis (probably non-R SB cards).
Note that lists with 4 Ports performed slightly above average, which is still a very good result, given the good overall performance of the lists in the dataset.

* I don’t think that the differences in win-% is significant between varying numbers of Piledrivers. The only thing that we can safely conclude here is, that cutting Piledriver is reasonable, as long as you somehow “make up for it”. Example: Lists with 0 Piledrivers were almost exclusively lists with Thalia.

* The difference in win-% between lists with 3 and 4 Mogg War Marshal is “large enough” to draw the conclusion that having 4 copies of MWM is strictly better than playing 3.

* Since list with 1 copy of Stingscourger are the only ones standing out among others, I conclude that 1 is the correct number, all other things being equal.
The same holds for 4 Gempalm Incinerator opposed to <4 copies and also for 1 Shatter-Goblin vs. 0 Shatter-Goblins.

* Skirk Prospector and Sharpshooter lists should be discussed in the same manner, as they frequently co-occur in decklists. It is therefore not surprising that they have the same pattern: “1” is slightly better than “0”. However, I’m hesitant to draw extreme conclusions here, as the the differences are only very small.

* I explored the 0-Chieftain-lists and the 1-Chieftain-lists further.
The exploration revealed that the successful 0-Chieftain lists are multicolored, with <4 Rishadan Port, <4 Piledrivers and with a full playset of both MWM and GIncinerator.
Lists with 1 Chieftains show no special features. It seems as if almost any list with exactly 1 Chieftain is more successful than other lists, irrespective of the varingy number of other cards in the MD. I think I can safely conclude that running1 Goblin Chieftain is a very good addition for the deck, but playing none does not hurt either. Thing get hairy when more than 1 Chieftain is involved though.

* Thalia, deserves to be discussed too. The way I presented the results it seems as if Thalia is making the deck worse. This is wrong, however. There were 28 records with 4 Thalias in MD, which performed way above average. However I decided to not presnt this information in the summary above for two reasons. First, N=28 is relatively low compared to other lists in comparison, which means that results are more likely to be random. Second, as far as I can remember, there is only one member on this forum, from which I got records of decklists with 4 Thalias in MD. This means that the very good performance can be a result of the skill of that player, rather than the value of the white creature card. My conclusion is that playing without Thalia in the MD is reasonable, but it would be too quick to conclude that she hurts the performance of the deck.

* I can’t tell yet which split between Krenko and SGC is “the best”. The successrates of the combinations in the summary are only slightly below average, which is in most cases a too small difference to be meaningful. It seems that the 1/2-split makes the deck better, however, this result is based on only 33 records. Unfortunately more testing is needed.

Next we take a look at how the number of CC2 spells affects the performance of the deck and how and if we can extract stereotype decklists out of that (just like last time).
In the scope of my analysis were 5 kinds of decklists, namely those with 10, 11, 12, 13 or 14 spells with manacost 2, which includes the following: Goblin Piledriver, Mogg War Marshal, Warren Instigator, Thalia, Stingscourger, Gempalm Incinerator. The analysis revealed that there are 3 lists that perform above average (to varying degrees: 11, 13 and 14), 1 list that is about equal to the average (10) and 1 list that scores below 62,77% (12). On the first glance the results partly replicate my results from my former analysis, however, there are a few nuances to them that I will explain in the following section.
Just as last time, I will try to sketch a stereotype decklist, whenever possible, starting with 14.

The 14er list aka. the THALIA build
Success-rate: 67,15%
N: 47
Stereotype-decklist
4 Wasteland, 4 Cavern of Souls
4 Rishadan Port
1 Plateau
3-4 Mountain
0-1 Karakas
5-6 Fetchlands
4 Aether Vial, 4 Goblin Lackey, 4 Goblin Warchief, 4 Goblin Matron, 4 Ringleader
2 Goblin Piledriver
4 Mogg War Marshal
4 Gempalm Incinerator
1 Stingscourger
3 Thalia, Guardian of Thraben
1 Krenko, Mob Boss
1 Siege-Gang Commander

The 13er list aka. the MonoR CLASSIC build
Success-rate: 66,16%
N: 68
Stereotype-decklist
4 Wasteland, 4 Cavern of Souls
4 Rishadan Port
10 Mountain
4 Aether Vial, 4 Goblin Lackey, 4 Goblin Warchief, 4 Goblin Matron, 4 Ringleader
4 Goblin Piledriver
4 Mogg War Marshal
1 Gempalm Incinerator
1 Stingscourger
1 Siege-Gang Commander

The 12er list aka. the Rg CLASSIC build
Success-rate: 60,29%
N: 95
Stereotype-decklist
4 Wasteland, 4 Cavern of Souls
2-3 Rishadan Port
1 Taiga
4-6 Mountain
5-7 Fetchlands
4 Aether Vial, 4 Goblin Lackey, 4 Goblin Warchief, 4 Goblin Matron, 4 Ringleader
4 Goblin Piledriver
3 Mogg War Marshal
4 Gempalm Incinerator
1 Stingscourger
1 Goblin Chieftain
1 Krenko, Mob Boss
1 Siege-Gang Commander

The 11er list aka. Undefined
Success-rate: 63,17%
N: 78
Stereotype-decklist
4 Wasteland, 4 Cavern of Souls
3 Rishadan Port
2 Duallands
3-4 Mountain
0-1 Karakas
5-6 Fetchlands
4 Aether Vial, 4 Goblin Lackey, 4 Goblin Warchief, 4 Goblin Matron, 4 Ringleader
0 Goblin Piledriver
3 Mogg War Marshal
3 Gempalm Incinerator
1 Stingscourger
4 Thalia, Guardian of Thraben
1 Krenko, Mob Boss
1 Siege-Gang Commander
1-3 Goblin Chieftain

The 14er list aka. the THALIA build
Success-rate: 67,15%
N: 47
Stereotype-decklist
4 Wasteland, 4 Cavern of Souls
4 Rishadan Port
1 Plateau
3-4 Mountain
0-2 Karakas
5-6 Fetchlands
4 Aether Vial, 4 Goblin Lackey, 4 Goblin Warchief, 4 Goblin Matron, 4 Ringleader
2 Goblin Piledriver
4 Mogg War Marshal
4 Gempalm Incinerator
1 Stingscourger
3 Thalia, Guardian of Thraben
1 Krenko, Mob Boss
1 Siege-Gang Commander

The 10er list aka. the WINSTIGATOR build??
Success-rate: 62,52%
N: 61
Stereotype-decklist
Not available.

Interpretation
The explanation of the first and the second list is pretty straightforward. Once again the data revealed that 14 cmc=2 spells are a typical trademark of a Rw THALIA decklist, which scores +4,38% above average. This is equally true for the “13” list as 13 cmc=2 spells seems to be defining for a MonoR CLASSIC decklist.
It’s a little more tricky to define the remaining lists properly. On the first glance, the “12” category appears to be a Rg version of the CLASSIC decklist. However, this is not in line with the results of the single-card analysis, which showed that both duallands and the combination of 2 Ports with duallands is an indicator of success.
The “11” list appears to be a variant of a THALIA build, probably with more duallands and 0 Piledrivers in favor of Chieftains. However, I’m unable to draw picture of that decklists that is as precise as the stereotype decklists of the first two categories.
The “10” list is equally problematic. Back when I did my last analysis I defined it as being a stereotype WInstigator list. This is not the case in today’s analysis. Only 40% of this sub-category were lists that featured some number of Warren Instigator, which means that most of the lists do in fact NOT deserve the WInstigator-label.
My overall interpretation is the following:
“14” is equal to THALIA build is equal to success.
“13” is equal to MonoR CLASSIC build is equal to success.
This relationship can’t be sketched for the remaining three categories. In fact it was foolish to assume that a stereotype decklist is EQUAL TO a statistical category, more precisely the number of CMC2 spells. I would rather say that it was more or less a matter of fortune that the “13” and “14” category correlate with a real deck subtype.

Last, some members indirectly asked me to look at how the number of “things to do by turn 2” affects performance, and this is what I did.
As a reminder, THIS is what I consider to belong in this category: “things to do by turn 2: any card that can be used by turn 2 without the help of Goblin lackey, Warchief, or Chrome Mox, which is: Goblin Lackey, Aether Vial, Skirk Prospector, Lightning Bolt, Tarfire, Gempalm Incinerator, Stingscourger, Thalia, Mogg War Marshal, Warren Instigator, Wasteland, Rishadan Port.
Additionally I checked the following things: number of 2-drops, number of cc1/3/4/5 spells, and average casting cost of the deck

Results:
* there is no linear relation between N of things to do by T2 and success rate.
* there is a curve-linear relationship between N of 2-drops and success rate. Lists with 7-9 2 drops scored above average (+1,48% to +2,42%). Other list had either an insufficient amount of data or were below average.
* lists with 9 or 10 CMC=1 spells scored better than those with 8, 11 or 12
* there is a negative linear relationship between number of CMC=3 spells and performance: the more you play, the worse it gets. Range from 8 CMC3 spells (65,66%) to 11 spells (58,18%). However, lists with less than 11 CMC=3 spells all scored above average, making only 11 spells a risky enterprise.
* no linear relationship was found for N of CMC=4 or CMC=5


IV. Afterword
I want to give credits to everyone who write complete tournaments reports on a regular basis and everyone in general who helped me with my data-collection by sharing his/her results in a fashion that I can use them for my analysis.
At the end of the day there’s not much left to say. Please be hesitant to rely on the stastistics too much. They can, by times, help you to get a new, fresh look on things, but they won’t help you becoming a better player. After all, statistics are worthless without some thoughts of interpretation. I have proposed some interpretations of the data myself, however this does not mean that I answered every question correctly.

Thank you for reading and feel free to discuss and use the results of my today’s analysis for your own good.
Question are, as always, appreciated.

-GL

[magicmerl]

Thanks for this anaylsis GoboLord, this is fantastic.

Since i have a Thalia focus, this is the bits I zoomed in on:

Rishadan Port - Lists with either 2 (+3,9%) or 4 (+0,5%) Ports performed above average. I want to run better mana in my three colour deck. So I'll try going to 2 Ports
Dual lands – Lists with either 1 (+3,38%) or 3 (+4,49 %) Duallands performed above average. MonoR lists performed slightly below average (-0,53%) I was already at 3 duals, so this is confirmation for me
Goblin Piledriver – Lists with 0 (+0,12%), 2 (+1,5%) or 4 (+1,05%) Piledrivers performed above average. Lists with 3 Piledrivers performed slightly below average (-0,17%). I'm not in love with PD, but my opponents tell me it's the card they fear the most.
Mogg War Marshal – Lists with 3 (+0,23%) or 4 (+2,52%) MWMs performed above average. So yeah, need these. Combined with PD this means that Thalia is prolly a SB card for me going forward
Thalia, Guardian of Thraben – Lists with 0 (+0,12%) Thalias performed better than those with 3 copies (-2,27%). Confirmed.

Gempalm Incinerator – Lists with 4 GIs performed above average (+0,41%). Lists with less than 4 copies performed below average (-0,24% to -1,05%). Confirmation of what I already suspected. Uncounterable cantripping removal is awesome vs creature based decks

Sharpshooter – Lists with 1 Skirk Prospector (+0,86%) performed slightly better than those with 0 copies (-1,23%).
Skirk Prospector - Lists with 1 Sharpshooter (+0,4%) performed slightly better than those with 0 copies (-0,55%)
Tin Street Hooligan or Tuktuk Scrapper – Lists with 1 “Shatter”-Goblin performed better (+2,06%) than those with 0 copies (-2,7%). So yeah, I think these 4 are locekd in as a MD tutor package

Krenko, Mob Boss and Siege-Gang Commander
0/2-split: -0,82% below avg
1/0-split: -1,52% below avg
1/1-split: -0,9% below avg
1/2-split: +3,14% above avg Really? That's a curious split for the 1/2 group. I'm constantly fighting to lower my curve, not beef it up. Can we see more of these 1/2 lists please?

* I don’t think that the differences in win-% is significant between varying numbers of Piledrivers. The only thing that we can safely conclude here is, that cutting Piledriver is reasonable, as long as you somehow “make up for it”. Example: Lists with 0 Piledrivers were almost exclusively lists with Thalia. Yeah, I think that Thalia is fighting directly with PD for a MD slot.

The exploration revealed that the successful 0-Chieftain lists are multicolored, with <4 Rishadan Port, <4 Piledrivers and with a full playset of both MWM and GIncinerator. Yep

* Thalia, deserves to be discussed too. The way I presented the results it seems as if Thalia is making the deck worse. This is wrong, however. There were 28 records with 4 Thalias in MD, which performed way above average. However I decided to not presnt this information in the summary above for two reasons. First, N=28 is relatively low compared to other lists in comparison, which means that results are more likely to be random. Second, as far as I can remember, there is only one member on this forum, from which I got records of decklists with 4 Thalias in MD. This means that the very good performance can be a result of the skill of that player, rather than the value of the white creature card. My conclusion is that playing without Thalia in the MD is reasonable, but it would be too quick to conclude that she hurts the performance of the deck. Yeah, I've been going a little sour on Thalia as a MD option recently. I think she can be a SB card for me.

* I can’t tell yet which split between Krenko and SGC is “the best”. The successrates of the combinations in the summary are only slightly below average, which is in most cases a too small difference to be meaningful. It seems that the 1/2-split makes the deck better, however, this result is based on only 33 records. Unfortunately more testing is needed. Agreed. I'd be inclined to think that the 1/2 split is a result of small sample size that isn't borne out by quadrupling the number of results

Results:
* there is no linear relation between N of things to do by T2 and success rate. Yeah, I think it's more the PROACTIVE things you do (i.e. permanents) that count
* lists with 9 or 10 CMC=1 spells scored better than those with 8, 11 or 12 So that means running Prospector OR Tarfire, right?
* there is a negative linear relationship between number of CMC=3 spells and performance: the more you play, the worse it gets. Range from 8 CMC3 spells (65,66%) to 11 spells (58,18%). However, lists with less than 11 CMC=3 spells all scored above average, making only 11 spells a risky enterprise. Goblin Chieftain....

Thank you for reading and feel free to discuss and use the results of my today’s analysis for your own good.
Question are, as always, appreciated.
Just want to say again, FANTASTIC. You do a great job keeping this thread alive and vibrant.

Based on the above results, here's the list that I'm planning on taking to my next tournament:
4 Cavern of Souls
4 Wasteland
2 Rishadan Port
4 Bloodstained Mire
3 Scalding Tarn
3 Mountain
2 Plateau
1 Taiga

4 Aether Vial
4 Goblin Lackey
4 Mogg War Marshal
4 Gempalm Incinerator
4 Goblin Warchief
4 Goblin Matron
4 Goblin Ringleader
3 Goblin Piledriver
2 Siege-Gang Commander
1 Skirk Prospector
1 Stingscourger
1 Tin Street Hooligan
1 Goblin Sharpshooter

Sideboard
3 Thalia, Guardian of Thraben
4 Chalice of the Void
4 Rest in Peace
3 Pyrokinesis
1 Boartusk Liege

That plan makes me a lot colder to combo decks G1, which means that I likely have to beat them once on the draw post board. Which is doable but not impossible. I plan on not being paired against Show and Tell or Energy Field decks. Other SB options I have considered: Krosan Grip, Angel of Despair, Mindbreak Trap, Relic of Progenitus, Wear//Tear.

[goblinsplayer]

Wow, great analysis. How do you feel about pyrokinesis in the maindeck? I see some people are running it and I am wondering how good it is in a star city games meta.

[Asgar]

Hey :)
Tranks for that Great analysis!
There's only 1 Thing i am Missing: the number of Lands played.
I could make some changes in my List Based in the data, but i would like to know, if 22-23 is still the correct number and to wich win-percentage the 21, 22, 23, 24 Lands lead.

Best Greetz. max

[ScatmanX]

I'm not sure I agree with that logic. Because of the presence of Goblin Matron we effectively have 5 copies of TSH. Going from 5 copies up to 6 copies is a pretty marginal increase that comes at the expense of a precious SB slot.

Just pointing out that, when you naturally draw your TSH, which does happen from time ti time, and then they discard it (jund,bug,esper and deathblade all run quite a lot of discard spells, + liliana), then you went from 'virtually 5', to 'actually 0'. That's a pretty big advantage for running 'virtuall 5' to 'virtually 6'.

@GobboLord: Thanks for the analysis, again. It (plus jon's last list) made me want to play with goblins again, so I'll probably take a new list for a spin this week (and write a detailed report for you =p).

[jrw1985]

@Gobolord: Fantastic and interesting analysis. There is a lot of interesting analysis to consider there. The most striking point to me was the success of lists running only 2 Piledrivers. That inspired me to brew another list:

Core 20 (I think we can define those 20 as the Core now)
4 Gempalm
4 MWM
2 Piledriver
1 Stingscourger
2 Warren Weirding
2 Krenko
2 Cheiftain
1 Lightning Crafter

22 lands

I think I'll be trying this out here soon, along with another goblin-centric sideboard.

[GoboLord]

@magicmerl:
It's not surprising that you find yourself confirmed with the results of the analysis. As far as I can remember the analysis includes quite a number of your decklists & results which were quite positive overall. I.e. you contributed a great deal to the postive overall performance of the dataset.

@goblinsplayer:
I, personally find Pyrokinesis in the MD a very viable choice, something like 1 or 2 copies sounds optimal to me. I don't have data to support my gut-feeling though.

@Asgar
TOTAL WIN-%: 62,77%, N=409
19 lands: 61,23%, N=13
20 lands: 55,19%, N=21
21 lands: 61,79%, N=29
22 lands: 64,84%, N=192
23 lands: 61,41%, N=150
24 lands: 66,50%, N=4

@ScatmanX and jrw1985:
thx for the kudos. Working on this analysis also made me want to play goblins again. Not only that, I think us three should start brewing goblin-lists together again. I always appreciate your feedback and insights.

[ScatmanX]

TOTAL WIN-%: 62,77%, N=409
19 lands: 61,23%, N=13
20 lands: 55,19%, N=21
21 lands: 61,79%, N=29
22 lands: 64,84%, N=192
23 lands: 61,41%, N=150
24 lands: 66,50%, N=4

I know that N=4 is too low for it to actually mean anything, but 66,50% is a quite high percentage. Maybe it tells us something, or with a higher number of samples the % should decrease, and spike at 22?

@ScatmanX and jrw1985:
thx for the kudos. Working on this analysis also made me want to play goblins again. Not only that, I think us three should start brewing goblin-lists together again. I always appreciate your feedback and insights.

Will pm my last list tonight along with my thoughts.

[magicmerl]

@magicmerl:
It's not surprising that you find yourself confirmed with the results of the analysis. As far as I can remember the analysis includes quite a number of your decklists & results which were quite positive overall. I.e. you contributed a great deal to the postive overall performance of the dataset.

Heh, yeah that is a problem. Additionally, there are a number of people who are netdecking Max Tietze's lists to good results. Possibly that's because he's found the best list. But possibly it's just the netdeck effect.

@Asgar
TOTAL WIN-%: 62,77%, N=409
19 lands: 61,23%, N=13
20 lands: 55,19%, N=21
21 lands: 61,79%, N=29
22 lands: 64,84%, N=192
23 lands: 61,41%, N=150
24 lands: 66,50%, N=4

As others have noted, the 24 lands list has the higest average but the lowest N. One thing I have seen from other games is that rather than use the average, use the average - 3*SD, which will give you a 99% confidence that the list is BETTER than. This will appropriately drop the high variance results.

Is that easy for you to do?

I know that N=4 is too low for it to actually mean anything, but 66,50% is a quite high percentage. Maybe it tells us something, or with a higher number of samples the % should decrease, and spike at 22?

I think that it will definitely regress to the mean with a larger sample size. Like I said above, I think that a better way of factoring in the uncertainty of the small sample size is to subtract a multiple of the standard deviation of each sample out. That will give us lower numbers, but more confidence in the results.

Just pointing out that, when you naturally draw your TSH, which does happen from time ti time, and then they discard it (jund,bug,esper and deathblade all run quite a lot of discard spells, + liliana), then you went from 'virtually 5', to 'actually 0'. That's a pretty big advantage for running 'virtuall 5' to 'virtually 6'.

I'm going to guess this has happened to you? Anyway, I wouldn't characterise it as 'a pretty big advantage'. I don't have an enormous need for artifact removal in my games, since typically the best artifacts are equipment, which I can handle by preventing combat damage or killing the equipped creature.

[LeoCop 90]

Of the cards you list that are a problem, graveyard hate nerfs DRS, Goyf and KotR. Thoughtseize, Abrupt Decay, Maelstron Pulse and Vindicate are just 1-for-1 answers that we don't care about because we have Ringleader/SGC. Liliana is bad because we have haste. Hymn is a little annoying but only one card.

So from my perspective I'd go with graveyard hate to strategically nerf their win conditions, and play out the games as if they are about grindy attrition.

I actually think that we don't care of 1-for-1 creature removals like swords to plowshares or lightning bolt. The question is different when this answer can destroy our vial (pulse, decay and vindicate) or even our lands (vindicate), or discard from our hand matron and ringleader (thoughtseize,hymn). All of these cards can ruin our gameplan, making us play like a fair deck (without matron/ringleader and mana to support them). And it is very difficult to win against goyfs and reliquaries if you can't develop card and board advantage.

About graveyard hate : the fact is that all the cards in my sideboard could be relevant against Junk : obviously pyrokinesis, then krosan grip for equipments, and relic as you stated. Last time I opted to board in only pyrokinesis.... I don't want to draw relic and then get killed by a jitte, or draw krosan grip and then get killed by big fatties. Next time i will try to board in relic because it hurts a lot of junk's cards and, most important, it replaces itself when cracked.
In your opinion, should I only side in relic, or relic + pyrokinesis ?

I don't have an enormous need for artifact removal in my games, since typically the best artifacts are equipment, which I can handle by preventing combat damage or killing the equipped creature.

I don't know how you can beat equipments so easily. I know we can kill equipped creatures with removal or use prospector, but it's not like we are a combo deck that wins the turn after. We don't have infinite removal spells, and also sacrificing a goblin per turn hurts our board position, not to mention that prospector can get killed.These plans sometimes may work, but sometimes may not.

Just pointing out that, when you naturally draw your TSH, which does happen from time ti time, and then they discard it (jund,bug,esper and deathblade all run quite a lot of discard spells, + liliana), then you went from 'virtually 5', to 'actually 0'. That's a pretty big advantage for running 'virtuall 5' to 'virtually 6'.

This happened to me once. Anyway I agree with magicmerl that a second tin street hooligan in side is not optimal.

Hey there fellow Warchiefs, Goblin-enthousiasts and lovers of statistics!

Fantastic analysis! it's good to see that you work so much on this thread, and I'm very interested in everything regarding goblins, although it's difficult to tell how much numbers and statistics matter.

[andrebonotto]

Hey there fellow Warchiefs, Goblin-enthousiasts and lovers of statistics!
The present article is about a statistical approach to find out what makes Vial-Goblin decklists successful in legacy.
(...)
Thank you for reading and feel free to discuss and use the results of my today’s analysis for your own good.
(...)

I'm just passing by to be one more to congratulate GoboLord for his (again) awesome work!

(I don't play Goblins for a while, but I try to keep up to date with this thread for its quality is pretty high.)

Sincerely,

- André

[fimo]

Thank you GoboLord for this awsome work!
I will for the sake of statistic play with a list that combines the highest win percentages from your analysis:

Rishadan Port - Lists with either 2 (+3,9%) or 4 (+0,5%) Ports performed above average. here we go with 2
Dual lands – Lists with either 1 (+3,38%) or 3 (+4,49 %) Duallands performed above average. MonoR lists performed slightly below average (-0,53%). I will splash a color
Goblin Piledriver – Lists with 0 (+0,12%), 2 (+1,5%) or 4 (+1,05%) Piledrivers performed above average. Lists with 3 Piledrivers performed slightly below average (-0,17%). 2 piledrivers
Mogg War Marshal – Lists with 3 (+0,23%) or 4 (+2,52%) MWMs performed above average. 4 MWM
Stingscourger – Lists with 1 Stingscourger performed above average (+0,53%). Lists with more or less than 1 Stingscourger performed below average. 1 of this guy
Gempalm Incinerator – Lists with 4 GIs performed above average (+0,41%). Lists with less than 4 copies performed below average (-0,24% to -1,05%). 4 of this guy
Goblin Chieftain – Lists with either 0 (+2,16%) , 1 (+3,78) or 4 (+0,95%) Chieftains performed above average. 1 of this guy
Sharpshooter – Lists with 1 Skirk Prospector (+0,86%) performed slightly better than those with 0 copies (-1,23%). 1 of this guy
Skirk Prospector - Lists with 1 Sharpshooter (+0,4%) performed slightly better than those with 0 copies (-0,55%) 1 of this guy
Tin Street Hooligan or Tuktuk Scrapper – Lists with 1 “Shatter”-Goblin performed better (+2,06%) than those with 0 copies (-2,7%). 1 of this guy
Thalia, Guardian of Thraben – Lists with 0 (+0,12%) Thalias performed better than those with 3 copies (-2,27%). here it is not clear as you write later on, but I will play 0 anyway
Krenko, Mob Boss and Siege-Gang Commander
0/2-split: -0,82% below avg
1/0-split: -1,52% below avg
1/1-split: -0,9% below avg
1/2-split: +3,14% above avg
There was a insufficient number of cases for all other possible splits. 1/2-split

Results:
* there is no linear relation between N of things to do by T2 and success rate.
* there is a curve-linear relationship between N of 2-drops and success rate. Lists with 7-9 2 drops scored above average (+1,48% to +2,42%). Other list had either an insufficient amount of data or were below average. Ok so I will keep the 2 drops between 7-9
* lists with 9 or 10 CMC=1 spells scored better than those with 8, 11 or 12. Ok so 9-10 1cc spells
* there is a negative linear relationship between number of CMC=3 spells and performance: the more you play, the worse it gets. Range from 8 CMC3 spells (65,66%) to 11 spells (58,18%). However, lists with less than 11 CMC=3 spells all scored above average, making only 11 spells a risky enterprise. I am very confused here. 1 chieftain is a good choice, 1 sharpsooter is a good choice, 4 warchiefs and 4 matrons are untouchable, it makes 10 3cc. How can 8 3cc have the highest performances if 1 chieftain and 1 sharpshooter boost the win% ???????????? can somebody help me interpreting this?

TOTAL WIN-%: 62,77%, N=409
19 lands: 61,23%, N=13
20 lands: 55,19%, N=21
21 lands: 61,79%, N=29
22 lands: 64,84%, N=192
23 lands: 61,41%, N=150
24 lands: 66,50%, N=4
I read this as it is correct to play many lands, between 22-24. I' ll go with 23

soooo.... based on these numbers ALONE the list should look like this:

4 Cavern of Souls
4 Wasteland
2 Rishadan Port
7 fetches
3 Mountain
3 DUALS.

4 Aether Vial
4 Goblin Lackey
1 Skirk Prospector
(9 1cc)

4 Mogg War Marshal
2 Goblin Piledriver
1 thalia or WI????????
(7 2cc)

1 TSH
1 Stingscourger

4 Gempalm Incinerator

4 Goblin Warchief
4 Goblin Matron
1 Goblin Sharpshooter
1 chieftain

4 Goblin Ringleaders

1 krenko
2 Siege-Gang Commander
(1:2 split)

It is hard to cover all requisites... Maximizing the win% I end up with 62 cards...should it just be be it? Actually to make the 3cc thing fit I should shave 1 matron and 1 warchief which sounds horrible but again, it is ust for the sake of statistics. It would bring the number of 3cc down to 8 which is the highest win%. Any input is appreciated but remember that I would like to stick to the highest numbers that GoboLord presented. I don t expect to win the next tournament with it, I just want to see how it works....
I ll let you know how it plays.

[ThoSha]

Hey goblin fellas, long time no see!

I was throwing back a goblin deck together and wanted to have some opinions on it.
Not sure about the numbers yet, but I'd like to cut Siegegang-Commander completely with running 8 Lords and 2 Krenkos.

It looks like this atm:

4 Cavern of Souls
10 Mountains
2 Rishadan Port
2 Chrome Mox
4 Wasteland

4 Aether Vial
4 Goblin Lackey
4 Mogg War Marshall
2 Goblin Piledriver
2 Warren Instigator
1 Stingscourger
3 Gempalm Incinerator
4 Goblin Warchief
4 Goblin Chieftain
4 Goblin Matron
4 Goblin Ringleader
2 Krenko, Mob Boss

SB:
Not sure yet

I know cutting Piledriver seems lame, but the swarmplan seems to make up for it.
Also I cant understand why noone runs WInstigator lately, that card is such a bomb.
Thoughts?

[fimo]

Hey goblin fellas, long time no see!

I was throwing back a goblin deck together and wanted to have some opinions on it.
Not sure about the numbers yet, but I'd like to cut Siegegang-Commander completely with running 8 Lords and 2 Krenkos.

It looks like this atm:

4 Cavern of Souls
10 Mountains
2 Rishadan Port
2 Chrome Mox
4 Wasteland

4 Aether Vial
4 Goblin Lackey
4 Mogg War Marshall
2 Goblin Piledriver
2 Warren Instigator
1 Stingscourger
3 Gempalm Incinerator
4 Goblin Warchief
4 Goblin Chieftain
4 Goblin Matron
4 Goblin Ringleader
2 Krenko, Mob Boss

SB:
Not sure yet

I know cutting Piledriver seems lame, but the swarmplan seems to make up for it.

Cutting piledrivers is very legit, I can say that from testing a lot without them. Also, GoboLord statistics show that cutting piledrivers is a reasonable choice. You do have to make up for it with some other form of damage output and your 4 chieftain and 2 krenko will do that very well. I think your list looks pretty solid actually, play with it and let us know how it goes.

Also I cant understand why noone runs WInstigator lately, that card is such a bomb.

The reason is that it is slow and unreliable if you play it on turn 2. On top of that it is the worst topdeck in the world. You need to invest so much energy and resources to have it connect that it is likely better to play something else instead. There are, however, people that had success with it.

[magicmerl]

About graveyard hate : the fact is that all the cards in my sideboard could be relevant against Junk : obviously pyrokinesis, then krosan grip for equipments, and relic as you stated. Last time I opted to board in only pyrokinesis.... I don't want to draw relic and then get killed by a jitte, or draw krosan grip and then get killed by big fatties. Next time i will try to board in relic because it hurts a lot of junk's cards and, most important, it replaces itself when cracked.
In your opinion, should I only side in relic, or relic + pyrokinesis ?

I am unsure whether you should SB both or just one, since once you start sideboarding in a significant number of cards you dilute the goblin synergy. However, I am sure that Relic should be the first card you bring in. Why run Pyrokinesis if you're taking our incinerators for them, or worse, making the incinerators you leave in so weak they can't kill anything?

I don't know how you can beat equipments so easily. I know we can kill equipped creatures with removal or use prospector, but it's not like we are a combo deck that wins the turn after. We don't have infinite removal spells, and also sacrificing a goblin per turn hurts our board position, not to mention that prospector can get killed.These plans sometimes may work, but sometimes may not.

I'm aware that the results I'm reporting with goblins vs Stoneblade is not the same as what other people are reporting. I guess it's my playstyle? I'm at heart a fairly controlling player, so my goals are to out-tempo my opponent in terms of board position while also having roughly as many goblins in play as in my hand. I also play the matchup as if the only thing that matters is stopping their equipment from hitting me (and keeping cards in hand for the inevitable board wipe).

Hey goblin fellas, long time no see!

I was throwing back a goblin deck together and wanted to have some opinions on it.
Not sure about the numbers yet, but I'd like to cut Siegegang-Commander completely with running 8 Lords and 2 Krenkos.

I know cutting Piledriver seems lame, but the swarmplan seems to make up for it.
Also I cant understand why noone runs WInstigator lately, that card is such a bomb.
Thoughts?

4 Warchiefs is a LOT. Probably too many. Given how insanely explosive Instigator is, I think that there's value in running KikiJiki over the 4th Chieftain (since he gives you unlimited goblins once you start copying matron and ringleader, and also lets you use EOT tricks to turn 1 Instigator/Piledriver into 3. Your list looks like it wants to be a textbook Winstigator list, which wants to but cannot run 4 of MWM, PD and Winstigator. I personally think that Piledrivers and Winstigators are more important than MWM.

I think that your list would be improved by turning your 8 two drops into max PD and Winstigator, Krenko -> SGC, and the 4th Chieftain -> KikiJiki.

To chime in with an opinion why Winstigator isn't being played much, I tried the list a while back and while it's very explosive, it doesn't help us win vs our worst matchups, the overwhelming victories it gives us I suspect we would have won anyway (making it win-more), and I had several games that I lost due to the inherent card disadvantage of Chrome Mox meaning that I HAD to imprint something to cast spells, but once my opponent killed my Winstigator, I was completely out of gas.

[goblinsplayer]

Is it okay to cut stingscourger from the list entirely considering that mono blue omnitell is the show and tell variant of choice? Considering sneak and show is nowhere to be found in the metagame nowadays.

[magicmerl]

Is it okay to cut stingscourger from the list entirely considering that mono blue omnitell is the show and tell variant of choice? Considering sneak and show is nowhere to be found in the metagame nowadays.

No.

Stingscourger is the BEST one-of tutor target, ahead of TSH, Prospector and Sharpshooter. You don't always want him, but when you do, you really want him.

[jrw1985]

Hello All!

I attended the inaugural Legacy Tournament at a new Magic and Comics shop in Minneapolis, Mead Hall Games and Comics. 6 Players showed. 3 of them were on Storm. Guess who I ended up paired against each round?

The List

Core 20
4 Gempalm
1 Stingscourger
2 Warren Weirding
1 Lightning Crafter
4 MWM
2 Piledriver
2 Krenko
2 Chieftain

3 Badlands
5 Fetch
4 Mountain
4 Waste
2 Port
4 Cavern

SB
3 Relic of Progenitus
4 Earwig Squad
2 Piledriver
1 Sharpshooter
1 Skirk Prospector
1 Tuktuk Scrapper
3 Tarfire

R1 TES
G1 I won the roll and mulled to 6. He killed me.
G2 I sided in 2 Piledriver, 4 Earwig, Tuktuk, 3 Relic, Sharpshooter, and Skirk. I sided out 4 Gempalm, Stingscourger, 2 Weirding, 4 MWM, and Lightning Crafter. My opening 7 was 1 Badlands, Skirk, 2 Earwig, Sharpshooter, Ringleader, Matron. I kept because if I ripped a land I could T1 Skirk, T2 Earwig, and with another ripped land a T3 Earwig. Well I played T1 Skirk. He sculpted. I missed the T2 land drop. He made 16 Goblins. I ripped a land and sac'd Skirk for Sharpshooter. He hit me for 16. I gunned em down next turn and played a Piledriver I had topdecked. He did nothing. I hit him with Piledriver and Prowled in Earwig. He scooped.
G3 I kept a super slow hand with Vials and Mana denial. But he kept his 7 too and just killed me.
0-1

R2 TES
G1 We both mulled to 6 and he killed me.
Same sideboarding as before.
G2 I led with a Lackey. He Probed me twice but couldn't go off yet, so he sculpted. T2 I hit with Lackey into Warchief into Earwig squad off a Cavern. I looked through his deck and counted 1 Ad Nauseum, 1 Tendrils, 4 Infernal Tutor, and 3 Burning Wish. I know some TES decks run only 3 Wish, but my opponent has been playing this deck forever and he's pretty good with it, and I seem to recall good players on the Source writing about how running 4 Wish was much better than running 3. So i knew he probably had a Wish in hand. I counted 4 LED. I realized with Burning Wish in his hand he could fetch up any win condition from his sideboard, so I couldn't just strip the win cons. So the only options were Rituals, Artifacts, or Sculpt cards. Well, the best card to pair with a Wish is LED, so I pulled 3 LED from his deck. He durdled on his turn. On my next turn I ripped Relic and attacked him down to 7. Again, he didn't go off on his turn. Relic stopped him from using PiF or GY loops, and I won.
G3 He cast Ad Nauseum from 15 life, and he burned himself out.
1-1

R3 ANT
G1 I Win the roll. We both keep 7. Mine is Cavern, Mountain, Port, Lackey Krenko Ringleader Piledriver. I lead with lackey. He durdles. I attack him and cheat in Krenko and cast Lackey. He durdles and fetches twice. I attack him down to 13 with Lackey and Piledriver, cheat in a MWM off Lackey, play a Vial I drew, and leave mana up to Port him on his upkeep. He Probes me and his life drops to 11. He goes off with Ad Neauseum and hits 0 life. The 2nd game tonight where the storm player Ad Nauseum'd to death.
G2 I sideboarded the same except without bringing in Sharpshooter because ANT doesn't run Empty the Warrens. I kept a 7 with 2 Vials, Relic, 2 Fetch, Piledriver, Matron. I can't really recall how this game exactly went, which was a shame because it was very interesting. I believe at one point I Ported him on upkeep, but I had 0 clock on him, so he just went off with Ad Nauseum that turn, even though he had to play his third land and drain his mana pool to do so. But the coast was clear for him to go off and he could just kill me the next turn anyway. He had a really bad AN and went from 20 to 3, but he found a hand that would allow him to win his following turn. Like I said, I had 0 clock on board. What I did have though, was Relic. I played Relic the following turn with mana up. His hand he had put together the previous turn was not a hand that could win through a Relic. I was able to cast Piledriver, then Cast and Vial in MWM and Warchief for the W.
2-1

So the sideboard and the deck did alright against Storm tonight, but my opponents had some pretty lousy Ad nauseums. I often had the right card at the right time this night. Relic did well, as did Earwig and Cheiftain. But it seemed to me that one strong advantage of not running Trap or Chalice is that your Storm opponent still needs to play around it to a degree. Just because you don't run Trap doesn't mean your opponent will just Storm off blind at the first opportunity. They need to make sure the coast is clear first, but doing so costs them a turn here and there. Otherwise, they just lose the game to FoW and MbT. And just because you don't run Chalice doesn't mean they won't side in Abrupt Decays and Echoing Truth, cards which are pretty mediocre against a Goblin deck that wants more beatdown.

Oh, and the shop was cool. Check them out sometime if you ever hang out downtown.

[orcanmail]

To chime in with an opinion why Winstigator isn't being played much, I tried the list a while back and while it's very explosive, it doesn't help us win vs our worst matchups, the overwhelming victories it gives us I suspect we would have won anyway (making it win-more), and I had several games that I lost due to the inherent card disadvantage of Chrome Mox meaning that I HAD to imprint something to cast spells, but once my opponent killed my Winstigator, I was completely out of gas.[/QUOTE]

2 Winstigators and 4 lackeys dropping 2 SGC/1 Krenko/4 Ringleaders is not the only outlet, playing 4 chieftains with 3 MWM, 2 piledrivers and 3 tarfires allows for another way of quickly overwhelming the board.
Warren Instigators are also good v combo, and as we have 4 vials and 4 caverns, not to mention 4 wastelands, we are already good v counters, and removal in the form of 3 tarfires, 2 stingscourgers and 3 gempalms, with 3 pyrokenisis in the SB allow us to deal with aggro. 3 chalices, 4 traps and 2 thorns in SB, 2 SB shatter goblins and 1 SB boartusk liege v e.plague have all made mono red winstigator a powerful deck.
I am not out of gas if my instigators are dealt with, and the acceleration of 2 chrome mox has won me games. My 12 basics also give me a stable mana base for double red cc, so maybe winstigators are just being overlooked due to playstyle by many players who prefer the classic aggro control list over the winstigator aggro list.