I totally agree with the tribal singleton thing, i've used a 100cs soldier tribal deck to some success in casual, making it 60 would be much more difficult, but probably just as fun
Now that you mentioned it, I checked and saw that if your opponent in the game you're watching is online, his avatar shows up correctly. If the opponent in the replay is offline, then he becomes a Serra Angel. I didn't know that.
I get why I don't particularly care about the statistics here but I don't get why that makes his article invalid. Just because you or I have little use for the information clearly does not mean that is so for anyone else. Plenty of people here have found a use for his analysis and some have expressed enjoyment for the article itself. I don't see how he 'needs' to do anything other than what he has been doing.
One way to look at this is that the author is having a lot of fun doing analysis that is not very meaningful when the project could be much more informative. The article approach is interesting, the methodology and interpretation are incredibly rough.
From a statistical point of view the author needs to tighten up his methodology. He needs to report the hypothesis that are being tested. He needs to report the statistical test he is using. He needs to report his sample size and standard errors for the statistics he reports. As a reader I see his table of ranked statistics and wonder if other readers also see that although plated geopede and tuktuk grunts are statistically significant (using an unknown test and unknown hypothesis), at face value there is no evidence that a probability of 60 and 65 are different from one another. The magnitudes of the probabilities convey nothing if the author does not test the statistics against each other.
I am not sure about the issues of sample bias brought up by other readers but if I were in the authors shoes I'd try a different model. Maybe a logit or probit with indicator controls for mono vs. splash decks or concede / straight win could address some of the criticisms.
I'm not sure how I missed that, but it was 42% and just below meeting the threshold for 90% statistical significance. I will include it in the next article.
'These statistics aren't perfect so they're useless'
'You haven't done the analysis for us, so these statistics are useless'
Yes there will be some major flaws with the results, that doesn't mean they're useless, it just means that you need to do some work and think about why some cards came out the way they did. Also, arguments about cards not being played because of concessions are reasonably minor and would become even less significant as the sample size increases.
And yes, you'll need to do your own analysis, he gives some explanations, but you need to think about why cards have been so successful for yourself.
If you get to having 5 lands on the table and cast a 5 mana offensive spell without losing or conceding, your baseline chance of winning is not 50%, but something like 70%.
You could try and model that effect, and compensate for it, but that sounds pretty tricky, especially when you consider good players usually concede earlier than bad ones.
Simpler would be punt and split the cards up by mana cost and maybe role, instead of colour. Comparing 2 mana and 5 mana spells is just not like-for-like. Certainly, in sealed, which is what the data set is for, the real choice you make is 'what 2 mana spells should I play?', not 'what red spells should I pick?'. Choice of colour follows that evaluation of cards, when you select a mana curve, and decide which colours have the best set of cards fitting into that curve.
Hey everyone, thanks for a great time, I was a little taken back to get into the top 4 first time but I guess 8/8 trampling indestructible treefolk on turn 5 can be a bit of a hand full :).
I hope to be back every week I can with new and different tribes.
With the holidays it has been a little rough finding time to get on to play some magic. You pretty much nailed it as I heard about/joined the event last minute and fish was the only legal tribe I had. Thanks for running a fun event Shard and hopefully I'll be able to play again next week.
Well as a note that RRR card you duressed me of was my Warp World I had intended to cast the following turn. Those lightening Rifts Ate my elfs alive. I just could not get any to stick long enough to get the warp world off.
I think the main reason we keep playing the same people is more the fact that there is not a huge number of people and those spikes stick out so when you play them more then once you remember it. Win or Lose though its still a great time. See you all this weekend to meet new new tribes.
Wow. I knew everyone looked at mtgotrader's prices, but I didn't know that bots were literally using them by the number. Do they scrape the website everyday? They can't possibly ask infobot that many questions.
Tarmogoyf - $37, 5 in stock
*FOIL* Tarmogoyf - $55.50, out of stock
DO IT! DO IT!! Set the foil Goyf to $10! Cause massive chaos by declaring an arbitrary value on a commodity that you don't even have! "Gee, if I did have them, I sell them for $10!" That'll teach them to automate their prices. The only people who will get burned are the people who aren't using their own data and who also fully automate their price changes.
If I register 10 new accounts, and make them all post "BUYING UNDERGROUND SEA - 20" in the classifieds, could I move the price of the card with mere words? I don't have 10 computers.
Timothy Hunt plays FNM at the Monster's Den in Minneapolis MN and I have met him on numerous occations and a buddy of mine was with him in St. Louis when he took 4th place. It was funny to see a name I recognized in your article. I am playing Spread' Em right now in FNMs and love the deck. The frustration of the opponent when you turn all their lands to islands is fantastic. Keep up the good articles and I will keep reading.
No apologies needed man, this was free to read and i found it to be pretty enjoyable. Your description of what makes the cards good is appreciated for someone who isnt always so quick to see why a certain card is great.
Like any statistical analysis, you need to understand what your analysis actually is measuring-- I understand it, but in light of this guy and other's comments, it would be better if you stated your H0 and HA a little more clearly.
It would be very beneficial if you would post your full data... I'd love to redo your analysis and can think of a few more that might be valid.
You should post the P value for every card, as well as the N for each card. Most people won't care, but us statistics gurus will appreciate. :)
Dear anonymous:
The results are obviously helpful if you understand them. These results are better then opinion; they mean something very specific, in fact. And yeah, the conclusions you can draw from this analysis aren't a pick order. You find that data set that can reach the conclusion you want and I'll analyze it, though.
not showing that burst lightning is better than tuktuk grunts to a statistically significant level shows that the percentages being output cannot be taken at face value, you would have a much better time comparing cards of the same mana cost. tuktuk grunts is generally better than spire barrage in sealed might be a conclusion (in draft where monored is possible or green red with a couple harrows this changes of course) shatterskull's double red make the comparison to the bladetusk murky. plated geopede is the best 2 drop in red, ok we already knew that, burst lightning is the best 1 cc spell, ok we already knew that, goblin war paint is better than a number of the other 2 drops might be a conclusion, I can see that, since in sealed few people are playing blue, it's also a pretty solid card in general and the other 2 drops aren't that exciting, slaughter cry and magma cry are the best 3 drops in sealed (though this is misleading since you often wont play either till much later and won't reveal you have the combat trick if you know youre losing meaning it'll see play more often when it actually helps you win the game, and rift won't be played when you're low on lands as much sometimes remaining uncast due to hoping for a 4rth or 5th land for a 4 or 5 drop) are they better than ruinous or molten ravager in sealed? I think most people would agree with that. The main thing here is you need to adjust your conclusions to take into account what you know or at least stick to situations where you know more than less. Comparing tuktuk to burst is just silly with these raw numbers. People that have been drafting for a long time know a lot about why cards are good, if they smell a rat, chances are theres a reason.
Still like the series, I think some people are becoming too venomous in their responses and you have become too defensive in your articles and replies. You seem to be lending more and more weight to the efficacy of your raw results the further along you go, which makes sense, people like to thin their efforts are meaningful and important. I would simply urge you to strongly consider the large number of factors that impede clear understanding coming out of such data.
Brave the Elements is 100% main deck.
Protection from = cant be blocked by... an unblockable double striking dualist with a machette wins you games.
I'd probably look at playing ony 2 Devout Lightcatsers main deck though.
as always excellent article. looks as though there was no wrong way to invest in extended so long as you did =)
I totally agree with the tribal singleton thing, i've used a 100cs soldier tribal deck to some success in casual, making it 60 would be much more difficult, but probably just as fun
Now that you mentioned it, I checked and saw that if your opponent in the game you're watching is online, his avatar shows up correctly. If the opponent in the replay is offline, then he becomes a Serra Angel. I didn't know that.
Well, I learned something today, thanks.
LE
I get why I don't particularly care about the statistics here but I don't get why that makes his article invalid. Just because you or I have little use for the information clearly does not mean that is so for anyone else. Plenty of people here have found a use for his analysis and some have expressed enjoyment for the article itself. I don't see how he 'needs' to do anything other than what he has been doing.
I just read Oath of Mages, I like the way they've balanced that cycle :P
@Lord Ermans ninja problem; I believe everyone is always shown as Serra Angels when watching a replay.
Apart from that another great tribal article :)
One way to look at this is that the author is having a lot of fun doing analysis that is not very meaningful when the project could be much more informative. The article approach is interesting, the methodology and interpretation are incredibly rough.
From a statistical point of view the author needs to tighten up his methodology. He needs to report the hypothesis that are being tested. He needs to report the statistical test he is using. He needs to report his sample size and standard errors for the statistics he reports. As a reader I see his table of ranked statistics and wonder if other readers also see that although plated geopede and tuktuk grunts are statistically significant (using an unknown test and unknown hypothesis), at face value there is no evidence that a probability of 60 and 65 are different from one another. The magnitudes of the probabilities convey nothing if the author does not test the statistics against each other.
I am not sure about the issues of sample bias brought up by other readers but if I were in the authors shoes I'd try a different model. Maybe a logit or probit with indicator controls for mono vs. splash decks or concede / straight win could address some of the criticisms.
I'm not sure how I missed that, but it was 42% and just below meeting the threshold for 90% statistical significance. I will include it in the next article.
Ok this argument basically seems to come down to:
'These statistics aren't perfect so they're useless'
'You haven't done the analysis for us, so these statistics are useless'
Yes there will be some major flaws with the results, that doesn't mean they're useless, it just means that you need to do some work and think about why some cards came out the way they did. Also, arguments about cards not being played because of concessions are reasonably minor and would become even less significant as the sample size increases.
And yes, you'll need to do your own analysis, he gives some explanations, but you need to think about why cards have been so successful for yourself.
where is goblin bushwacker ?
If you get to having 5 lands on the table and cast a 5 mana offensive spell without losing or conceding, your baseline chance of winning is not 50%, but something like 70%.
You could try and model that effect, and compensate for it, but that sounds pretty tricky, especially when you consider good players usually concede earlier than bad ones.
Simpler would be punt and split the cards up by mana cost and maybe role, instead of colour. Comparing 2 mana and 5 mana spells is just not like-for-like. Certainly, in sealed, which is what the data set is for, the real choice you make is 'what 2 mana spells should I play?', not 'what red spells should I pick?'. Choice of colour follows that evaluation of cards, when you select a mana curve, and decide which colours have the best set of cards fitting into that curve.
Hey everyone, thanks for a great time, I was a little taken back to get into the top 4 first time but I guess 8/8 trampling indestructible treefolk on turn 5 can be a bit of a hand full :).
I hope to be back every week I can with new and different tribes.
With the holidays it has been a little rough finding time to get on to play some magic. You pretty much nailed it as I heard about/joined the event last minute and fish was the only legal tribe I had. Thanks for running a fun event Shard and hopefully I'll be able to play again next week.
lol yes new tribes galore hopefully. I know i have three different decks ive built in anticipaion.
That's my "red mana" post. I didn't realize I wasn't logged in.
Actually, paragraph 3 is definitely red, but paragraph 4 is very blue.
Well as a note that RRR card you duressed me of was my Warp World I had intended to cast the following turn. Those lightening Rifts Ate my elfs alive. I just could not get any to stick long enough to get the warp world off.
I think the main reason we keep playing the same people is more the fact that there is not a huge number of people and those spikes stick out so when you play them more then once you remember it. Win or Lose though its still a great time. See you all this weekend to meet new new tribes.
Wow. I knew everyone looked at mtgotrader's prices, but I didn't know that bots were literally using them by the number. Do they scrape the website everyday? They can't possibly ask infobot that many questions.
Tarmogoyf - $37, 5 in stock
*FOIL* Tarmogoyf - $55.50, out of stock
DO IT! DO IT!! Set the foil Goyf to $10! Cause massive chaos by declaring an arbitrary value on a commodity that you don't even have! "Gee, if I did have them, I sell them for $10!" That'll teach them to automate their prices. The only people who will get burned are the people who aren't using their own data and who also fully automate their price changes.
If I register 10 new accounts, and make them all post "BUYING UNDERGROUND SEA - 20" in the classifieds, could I move the price of the card with mere words? I don't have 10 computers.
Timothy Hunt plays FNM at the Monster's Den in Minneapolis MN and I have met him on numerous occations and a buddy of mine was with him in St. Louis when he took 4th place. It was funny to see a name I recognized in your article. I am playing Spread' Em right now in FNMs and love the deck. The frustration of the opponent when you turn all their lands to islands is fantastic. Keep up the good articles and I will keep reading.
No apologies needed man, this was free to read and i found it to be pretty enjoyable. Your description of what makes the cards good is appreciated for someone who isnt always so quick to see why a certain card is great.
William,
You rock.
Love this series.
Like any statistical analysis, you need to understand what your analysis actually is measuring-- I understand it, but in light of this guy and other's comments, it would be better if you stated your H0 and HA a little more clearly.
It would be very beneficial if you would post your full data... I'd love to redo your analysis and can think of a few more that might be valid.
You should post the P value for every card, as well as the N for each card. Most people won't care, but us statistics gurus will appreciate. :)
Dear anonymous:
The results are obviously helpful if you understand them. These results are better then opinion; they mean something very specific, in fact. And yeah, the conclusions you can draw from this analysis aren't a pick order. You find that data set that can reach the conclusion you want and I'll analyze it, though.
not showing that burst lightning is better than tuktuk grunts to a statistically significant level shows that the percentages being output cannot be taken at face value, you would have a much better time comparing cards of the same mana cost. tuktuk grunts is generally better than spire barrage in sealed might be a conclusion (in draft where monored is possible or green red with a couple harrows this changes of course) shatterskull's double red make the comparison to the bladetusk murky. plated geopede is the best 2 drop in red, ok we already knew that, burst lightning is the best 1 cc spell, ok we already knew that, goblin war paint is better than a number of the other 2 drops might be a conclusion, I can see that, since in sealed few people are playing blue, it's also a pretty solid card in general and the other 2 drops aren't that exciting, slaughter cry and magma cry are the best 3 drops in sealed (though this is misleading since you often wont play either till much later and won't reveal you have the combat trick if you know youre losing meaning it'll see play more often when it actually helps you win the game, and rift won't be played when you're low on lands as much sometimes remaining uncast due to hoping for a 4rth or 5th land for a 4 or 5 drop) are they better than ruinous or molten ravager in sealed? I think most people would agree with that. The main thing here is you need to adjust your conclusions to take into account what you know or at least stick to situations where you know more than less. Comparing tuktuk to burst is just silly with these raw numbers. People that have been drafting for a long time know a lot about why cards are good, if they smell a rat, chances are theres a reason.
Still like the series, I think some people are becoming too venomous in their responses and you have become too defensive in your articles and replies. You seem to be lending more and more weight to the efficacy of your raw results the further along you go, which makes sense, people like to thin their efforts are meaningful and important. I would simply urge you to strongly consider the large number of factors that impede clear understanding coming out of such data.
What is the point of these articles?
Duplicate posting.
Quote: And in one game when my opponent had Sphinx of the Steel Wind AND Iona, Shield of Emeria in play, I found myself a Blazing Archon.
True story. :( I believe he was at like 4 or 5 with no fliers.