Probability of Tournament Succes Based on Supplied Data

[Valhallan42nd]

Hello all, you might remember me from the Statistics Project Thread: [url]http://www.lounge.belloflostsouls.net/showthread.php?t=2659[/url]

I'm here to post my finding on the data set I had when I turned in the report.

What is the better predictor of tournament success: the army you play, years playing the army, years in the hobby, or population you’re from?

Most Warhammer 40,000 tournaments consist of three games. To be considered a top general, and therefore have a shot at winning the tournament, it is best to win all three of your games. Given the data provided by my various populations, I will attempt to ascertain the most consistent factor in successful generals, and project how often that factor propels them into the top tier (three wins) over a series of 25 simulated tournaments.

I generated a win percentage based on the factors expressed above. Each population will be divided in logical fashion, based on the following:

Army: each army will be a separate group. There were three groups that received two entries or less (Blood Angels, Black Templar, and Necrons). These populations will not be considered for portion of this simulation, as their data sample size is too small to be a good sample.

Years playing the army: the data will divided into 4 groups, based on their positions within the quartiles of this group’s data: 0.25-.75, .76-1.99, 2.00-5.49, and 5.50+

Years in the hobby: the data will divided into 4 groups, based on their positions within the quartiles of this group’s data: .25-2.99, 3.00-7.99, 8-13.99, 14+

Population: Data will be used from the three largest respondent groups, and all the smaller forums combined into a category marked “other”. I have decided to combine data from groups with less than 15 responses due to the small sample size skewing win percentages.

Random numbers were generated via MS Excel. I used these random numbers to compare them against the group’s win percentage as pulled from the data. If a number was found to be less then that group’s win percentage, then conditional formatting highlighted the random number entry. A series of three wins will be considered a success.

Here is what I found:

The greatest predictor of success in the data that I obtained was one’s army selection. I was not surprised by this, as this has been a consistent contention of the online community .To be honest, I was surprised at the army at the top of the heap, win percentage wise. Chaos Daemons are held by a great many “in the know” to be a mid-tier army due its struggle against the new paradigm of Mechanized armies (armies with the majority of it’s forces in armored personnel carriers with tank support). With a larger sample size, you might see some shifting of this percentage, as well as a shifting of the win percentage of Dark Angel armies. There was also a significant amount of evidence that poor army selection would be a severe hindrance in tournament success. Suffice to say, I won’t be likely to buy a Tau Empire army anytime soon. An army built around advanced long range weaponry, their lack of close combat ability, coupled with the increased effect of cover in 5th Edition has put a severe crimp in their ability to function as a successful army.



Although the simulations pointed next to the next highest success factor as being “8-13.75 Years in the Hobby”, I wanted to point out the consistency of those who post in online forums. Even the least of them (DakkaDakka) would achieve a top tier position over 16% of the time in my simulation, with the others ranging from 25% to 28% of the time. Getting into the top of your game a quarter of the time is certainly to be lauded. From my experience, the advice given on such forums is certainly a significant factor in forum browser’s consistent performance, and is to be recommended for those who wish to improve their tournament performance.

[Melissia]

The images really need to be larger. They were reduced in size, making it hard to tell which color represents what on the first graph, and making the words blurry on the second.


Also, this doesn't entirely seem that useful, certainly not to me. It doesn't take into account army build and so on. Choice of army is only the first step in creating a proper army list, and just within C:WH you can create one army that is extremely effective and another that is laughable (see the difference between Inquisition and Sororitas based army lists).


Furthermore, you should take into account percentage of wins in relation to the number of players that were present rather than percentage of the whole. Because the army choices are unevenly spread across the various codices, it doesn't really help much to see wins as a whole-- because those which are more frequently taken can quite easily have more wins overall even if they have a lower win percentage.


Warhammer 40,000 has a very complex set of statistics, and is not easy to measure. I salute you for trying, despite my own objections.

[PhoenixFlame]

Thanks for posting your results, interesting read

[Valhallan42nd]

The images really need to be larger. They were reduced in size, making it hard to tell which color represents what on the first graph, and making the words blurry on the second.


Also, this doesn't entirely seem that useful, certainly not to me. It doesn't take into account army build and so on. Choice of army is only the first step in creating a proper army list, and just within C:WH you can create one army that is extremely effective and another that is laughable (see the difference between Inquisition and Sororitas based army lists).

One does not go into war with the Data Set one wants, one goes into war with the Data Set one has. I see your points, and beleive me, I would have liked to study them further.But, to take into account Army Build (a nebulous term), I would have had to create an experiment with people all using the exact same template (otherwise it's a different Army Build). And then get a data pool or each of those templates in excess of 100. Which then doesn't elminate the factors inherant in generalship, and luck. And I'd have to repeat that for each template. I simply didn't have the budget for that. I felt it better to zoom up in scale and work with the data I could get for free, provided by you fine folks.

When I get my grant, I'll build a better study.

Furthermore, you should take into account percentage of wins in relation to the number of players that were present rather than percentage of the whole. Because the army choices are unevenly spread across the various codices, it doesn't really help much to see wins as a whole-- because those which are more frequently taken can quite easily have more wins overall even if they have a lower win percentage.

Warhammer 40,000 has a very complex set of statistics, and is not easy to measure. I salute you for trying, despite my own objections.

The simulation doesn't need to factor in the number of players in the tourney, it merely needs to measure the likelyhood of you winning three games. You never play against more than three opponents anyway. Now, I do understand the factors you mean: that, in a Swiss system, the top players play against other top players, so each game increases in difficulty, but there was no way of measuring that without significant data from tournament organizers. Which I wasn't going to get over the course of a week or two.

I understand the frustration from the study's limitations, but I think they're still a usefull guide for the top factors that influence tournament army success.

[Miggidy Mack]

I find it hilarious that Necrons aren't even on the chart, along with Blood Angels and Black Templar. Is that because they haven't won a tournament since 5th edition or do they just have such a small sample size that they aren't viable for the numbers.

You mentioned they only had 2 or so entires, I was just wondering why the entry amount was so low.

How would your numbers be skewed if, for instance, a bunch of bad players played Orks & Marines but also some incredibly skilled players? I believe this to be the case. I have yet to meet a Daemons player who didn't really know his way around controlling a games pacing.

[Lerra]

Dark Angels might have scored so well because DA players tend to be older and more experienced (from personal experience). Dark Angels were a very poplar army in the "old days", less so than they are now.

If you ever do this again (or if you need to write a paper on this ), a suggestion for next time would be to drop the codices and ask "Which of the following describes your army?" and have check boxes for Assault, Ranged, Mech, Horde, etc. Or ask specific questions like "How many weapons with the melta descriptor does your primary army list have?", "How many armored units does your primary army list have?" etc.