Jump to content
Sign in to follow this  
StoneAge

Will the Rumblings of War tournament (Row) be Returning

Recommended Posts

Greetings.

It's been a while since I've been here. So long that I have lost my Nabla account password and access to my old email address - hence the modified username. Incidentally I bought CMAK a couple of months ago and have now been playing it occasionally. Had quite a workload for a few years.

I am happy to see the tournaments back. The further development of the tournament system stopped when Treeburst155 suddenly went MIA; hope everything is fine with him. Now that we have an active tournament manager we can put the show on the road again.

The Nabla system was developed for ranking groups of players that all play against each other. Typically you'd have a number of players in different sections all playing against each other. You would rank the players within each section to choose those that proceed to the final round. In the final round, all those selected players would again play against each other to find the winner of the tournament.

I have understood that in the current tournament there would be no second round. If this is the case, then the current version of the system is probably not the best one, since it provides very little information for comparing players in different sections. Thus a modification to the system would be needed. This sounds like an opportunity to do some interesting math and coding, which is fine by me, but would take some time and effort. Therefore, I would appreciate it if WineCape could verify that a different system is really needed before I proceed any further.

What kind of a different system could we have? There are many possibilities. I want to make it clear at this point that I am currently not on a first name basis with any of these: while my background is in machine learning and computer science, I am not specialised in ranking systems. Anyway, as for the different possibilities, item response theory has been mentioned in this thread. It seems to have been developed for assessing capabilities of individuals based on tasks, not pairwise competition. Statistical models of pairwise comparison and other methods have been used to learn rankings, for example, in college football.

One option is to device a system heuristically based on the current system - as compared to a fully fledged generative probabilistic model and the associated estimation method. The first heuristic idea that came to my mind is the following. Let A > B denote the fact that A played better than B in any scenario in a tournament, where better is defined in terms of the median score. If A > B > C < D, then not only does A gain final points w.r.t. B but also w.r.t. C; however, the chaining would stop at D. The propagation of the score could be weighted in a geometrically decreasing manner: luck plays a role in the game and different scenarios test different skills. The chaining would be taken into account in scheduling the tournament.

This kind of a system would have its own side effects due to the assumptions in the system. For example, assume that B played seriously against C with the end result B > C, but would then lose interest with the end result A > B. These battles would imply A > C even though it might well be the case that A played worse than C. However, regardless of whether a heuristic method or a fully specified model is used, the fact is that this is a learning problem, in which we try to learn a ranking from pairwise game results, and you can not learn without assumptions. In a full generative model the assumptions are laid out explicitly so that everybody can see and discuss them. In a heuristic model the assumptions are included somewhere in the system. This does not necessarily mean that a completely probabilistically specified system would be better: it can be much more complicated, more difficult to understand and discuss, intractable (impossible to estimate) and, at the end, just as questionable when it comes to the underlying assumptions.

In any case, I am convinced that we will have an electrifying tournament with hopefully a less electrifying :) scoring system running for you soon.

Share this post


Link to post
Share on other sites
Greetings.

It's been a while since I've been here. So long that I have lost my Nabla account password and access to my old email address - hence the modified username. Incidentally I bought CMAK a couple of months ago and have now been playing it occasionally. Had quite a workload for a few years.

I am happy to see the tournaments back. The further development of the tournament system stopped when Treeburst155 suddenly went MIA; hope everything is fine with him. Now that we have an active tournament manager we can put the show on the road again.

The Nabla system was developed for ranking groups of players that all play against each other. Typically you'd have a number of players in different sections all playing against each other. You would rank the players within each section to choose those that proceed to the final round. In the final round, all those selected players would again play against each other to find the winner of the tournament.

I have understood that in the current tournament there would be no second round. If this is the case, then the current version of the system is probably not the best one, since it provides very little information for comparing players in different sections. Thus a modification to the system would be needed. This sounds like an opportunity to do some interesting math and coding, which is fine by me, but would take some time and effort. Therefore, I would appreciate it if WineCape could verify that a different system is really needed before I proceed any further.

What kind of a different system could we have? There are many possibilities. I want to make it clear at this point that I am currently not on a first name basis with any of these: while my background is in machine learning and computer science, I am not specialised in ranking systems. Anyway, as for the different possibilities, item response theory has been mentioned in this thread. It seems to have been developed for assessing capabilities of individuals based on tasks, not pairwise competition. Statistical models of pairwise comparison and other methods have been used to learn rankings, for example, in college football.

One option is to device a system heuristically based on the current system - as compared to a fully fledged generative probabilistic model and the associated estimation method. The first heuristic idea that came to my mind is the following. Let A > B denote the fact that A played better than B in any scenario in a tournament, where better is defined in terms of the median score. If A > B > C < D, then not only does A gain final points w.r.t. B but also w.r.t. C; however, the chaining would stop at D. The propagation of the score could be weighted in a geometrically decreasing manner: luck plays a role in the game and different scenarios test different skills. The chaining would be taken into account in scheduling the tournament.

This kind of a system would have its own side effects due to the assumptions in the system. For example, assume that B played seriously against C with the end result B > C, but would then lose interest with the end result A > B. These battles would imply A > C even though it might well be the case that A played worse than C. However, regardless of whether a heuristic method or a fully specified model is used, the fact is that this is a learning problem, in which we try to learn a ranking from pairwise game results, and you can not learn without assumptions. In a full generative model the assumptions are laid out explicitly so that everybody can see and discuss them. In a heuristic model the assumptions are included somewhere in the system. This does not necessarily mean that a completely probabilistically specified system would be better: it can be much more complicated, more difficult to understand and discuss, intractable (impossible to estimate) and, at the end, just as questionable when it comes to the underlying assumptions.

In any case, I am convinced that we will have an electrifying tournament with hopefully a less electrifying :) scoring system running for you soon.

A long quote, since you're usually quotable in toto. The "h" in your name don't fool us. Get working. :D OK Nablah, here's my take on the tourney format, being aware that changing the format might be not good, depending on your workload in re-programming NABLA to accommodate it:

Firstly, I would love to NOT have to have a knockout round. The tournament can still be 72 players total, but you will have 3 x 24-player mini tourneys within the tournament, each tourney having 3x8-player sections, with each player battling with 7 scenarios total. Given the underlying assumptions of the current NABLA, is this possible/desirable? The idea is not so much having a single winner out of all 72 players, but having 3 Tourney winners within the overarching Tournament. This would be the only change I would like to make to Nabla scoring/scheduling, if nothing else.

The issue is for Nabla to compare players accurately, that among the 9 x 8-player Sections total , you get 3 x SECTION winners WITHIN each Tourney. The best/highest section scored winner within EACH of TOURNEY I, II and III should get the prize as the overall 3 winners of the Tournament.

The reason is simple: No extra play-off scenarios are needed, and more importantly, not ALL losing participants in their respective sections, after losing out on qualifying for the play-off is previous RoW's necessarily want to play these extra play-off scenarios to SCORE them more accurately for the NABLA system for those involved in the play-offs. In the past, we at RoW had no issue in "cajoling" losing section players to do so. But it is not ideal. Can this be accommodated in NABLA? Thus, instead of having a sectioned round robin + a play-off format, you have only a round robin format with 3 winners.

For clarity sake:

A 'Tournament' = 72 players total [RoW VI]

A 'Tourney' = 24 players each within the Tournamet [named Tourney I, II, & III]

A 'Section' = 8 players each, playing 7 battle scenarios [can be #'ed sequentially, or just plainly "1-3" within each Tourney]

Secondly, and certainly a "less wanted feature" given the drastic change in mechanics of scoring to accommodate it, if we decide to do away with Section play altogether, is it even desirable/possible to lump all players in one single group/Tournament to play a "Round Robin" / a-la-Swiss style Chess tournaments, battling against SOME opponents only, and still declare the top 3 winners, despite these 3 not having played each in the tournament due to design? Maybe this method is just to much of a drastic change on the existing NABLA scoring/scheduling to bother about? Don't bother with scoring method if this if a pipe dream altogether. :)

I'm not worried really on the second method, but the first proposed method would certainly be a "wish" for any tournament director, or at least to have such an extra option within the NABLA scoring/scheduling system. That is, no knock-out phase.

Share this post


Link to post
Share on other sites

No, no! Given that my eyes started glazing over after the word "Greetings" from Nabla due to all the gobbledegook about random string theory, voyeristic models and procreation of the score I think Nabla with some blah included is perfect as it is.

;)

Regards

KR

Share this post


Link to post
Share on other sites

Here is my take on how the ROW could be run this time

1) create the 'Tournament' = 72 players total [RoW VI].*

2) Each player plays all 7 scenarios.*

3) Nabla scoring system used to work out axis mean score's and allied mean score's.*

4) Nabla Scores allocated on distance from mean.

5) Each players top and bottom scores dropped.*

6) final 5 scores added togeather and turned into a ladder.

Top player is the winner !!!

Woohoo start again.

*1 number players does not matter, long as it is even number of players.

*2 Split group into two groups group 1 gets 3 Axis 4 allied, group 2 get 4 Axis 3 allied scenario's

* this is from my understanding of how nabla worked in ROWIII, IV and V

*3 scoring system works out the average/mean of the scenario (eg axis 5 allied 95 is the average/mean result)

*3 scoring can also leave games out where a player dropped out.

*5 like olympic diving, removes walkover scores.

Don't know if this would work, it would help to stream line the tourney and give every body an idea where they fall in the CM:BN world.

Share this post


Link to post
Share on other sites

What kind of a different system could we have? ... Anyway, as for the different possibilities, item response theory has been mentioned in this thread. It seems to have been developed for assessing capabilities of individuals based on tasks, not pairwise competition. Statistical models of pairwise comparison and other methods have been used to learn rankings, for example, in college football.

One option is to device a system heuristically based on the current system - as compared to a fully fledged generative probabilistic model and the a....

Welcome back Nabla,

It seems I'm the only one here familiar with the use of Rasch statistics, but unfortunately I'm no mathematician so I'm not going to be discussing pros and cons of Rasch vs other possible statistical solutions.

Anyway, just to let you know that the Facets program that I use has been calibrated so that it runs with the kind of data that we get in CM games (pairwise/percentage), and that I trialed it with the previous RoW tournament, even though it was not an ideal set up (the players were only connected through their performances on the scenarios, not through being connected with each other), it gave very similar results to those obtained by the Nabla system.

I think using Facets would be ideal for a single tournament of 70 plus players or more, split into three divisions.

In terms of tournament design, if WineCape wants three divisions it is easy to do, but for connectivity (and fun) each player would have a couple or so cross division games with players from the other pools.

So, you might have say, three pools of 30 players with seven scenarios, you would play five against players of your own pool, and one against a player from each of the other pools. In this way Rasch should be able to compare achievement across the groups. Not only would you get a winner of each tournament, you would be able to select one overall champion as well.

However, as I've said before, I'd like to try it out on a smaller scale before doing it on RoW.

Share this post


Link to post
Share on other sites
When can the waiting list expect to know whether they're in or out and how? Via PM?
Timeframe of RoW VI as follows:

(1) Nablah reasoning of a tournament, which is a pair <N,A> where N is a finite set of at least two elements, which we will call players, and A is a zero-one matrix, such that its diagonal entries are zeroes and for all i, j ε N, i ≠ j, aij +aji = 1. The interpretation is that each pair of players in N played each other once, and aij = 1 if, and only if, i beat j.

(2) Nablah gets a bright spark, drops the "h" in his name, and produces another beautiful math equation for ranking/rating RoW in different sections. Failing that, repeat step (1);

(3) He writes a new front-end for NABLA, to be used by moi, the tourney director;

(4) Demo and CMBN by now released in wild. Everyone forgets NABLA and RoW for awhile;

(5) Scenario designers approached with concept paper for designing 7 scenarios specifically for RoW to test players abilities w/ every CM unit capable of being clicked w/ mouse pointer;

(6) Invited Scenario designers goes in evil genius lock-down mode and design diabolical battles, never seen before; the need to wear clothes while doing so not obligatory;

(7) Invites and dates set and announced on forum for tourney start. Vets that showed interest to play RoW VI will get 10-14 days grace to reply to invites, following that invites send to newbies, who showed interest in playing, all via PM.

(7a) (7) above not possible for Newbies without obligatory thesis titled: Why we should character assassinate your name upon drop-out;

(8) Great fun and gnashing of teeth to be had for +/-150 days, depending on state of intoxication while playing;

(9) Winners announced, I drink wine, send corks as prizes, winners fêted as the greatest ever to set foot on Normandy.

Share this post


Link to post
Share on other sites

Please put me on the potential invitee list.

I got in to the CM sport too late to have participated in any of the previous ROWs, but I am currently the top ranked CMAK player at We Band of Brothers.

Share this post


Link to post
Share on other sites
Firstly, I would love to NOT have to have a knockout round.

Ok.

The tournament can still be 72 players total, but you will have 3 x 24-player mini tourneys within the tournament, each tourney having 3x8-player sections, with each player battling with 7 scenarios total.

Dividing 72 players into 9 sections with 8 players in each section would basically give you 9 winners. Unless we assume that the different sections are equally strong.

Secondly, and certainly a "less wanted feature" given the drastic change in mechanics of scoring to accommodate it, if we decide to do away with Section play altogether, is it even desirable/possible to lump all players in one single group/Tournament to play a "Round Robin" / a-la-Swiss style Chess tournaments, battling against SOME opponents only, and still declare the top 3 winners, despite these 3 not having played each in the tournament due to design?

This is what is to be investigated next.

Would an adaptive scheduling policy be out of order? That is, at the beginning schedule only the first round (first scenario), then schedule the second round (second scenario) based on the results of the first round etc. This, combined with result propagation discussed above, might provide some fascinating possibilities. But I understand that it might be difficult to run such a tournament in a reasonable time due to the fact that PBEM can be so slow, especially with participants all over the world.

I will reply to the other responses later.

Share this post


Link to post
Share on other sites
Originally Posted by WineCape

Firstly, I would love to NOT have to have a knockout round.

Ok.

For clarity sake, this is meant as an OPTION to the current knock-out system Nabla is programmed for, not a replacement by the way.

Share this post


Link to post
Share on other sites
Would an adaptive scheduling policy be out of order? That is, at the beginning schedule only the first round (first scenario), then schedule the second round (second scenario) based on the results of the first round etc. This, combined with result propagation discussed above, might provide some fascinating possibilities. But I understand that it might be difficult to run such a tournament in a reasonable time due to the fact that PBEM can be so slow, especially with participants all over the world.
Aye. As per Swiss system style, used in Chess tournaments. It's usually used in big Open Championships were there a many a participant. This could be a solution, but unlike Grandmaster Chess where you have, say 40 moves in the first 2 hours, with a secondary time control after this, in PBEM, as you indicated, it will play havoc with schedules, even if you set a (rather short) 20-day limit for Scenario A to be completed by all participants.

In fact, this might be the ONLY way to schedule/rank all 72 players in one go and decide who meets who in the next round; those that have scored roughly the same amount of points/wins in previous round. Tie breaking systems, as used in chess, will then decide those winners who scored the same wins/points after all scenarios played. But scheduling a time limit to Scenarios are fraught with danger. Some play fast, some play slow, some can't send a single email move for x-amount of days due to Real Life. Unless we find an acceptable scheduling/time frame solution, or a way/solution to only pair match players for the next round AMONG those that have indeed finished their previous round.

The question, Nabla, is if the above Swiss based system, after looking at other possible methods, is in fact your ONLY choice to rank players effectively from #1-#72 when not all players have met all in combat. For we then don't have to invent the wheel, as Swiss play and their various tie-braking systems are very well documented.

Share this post


Link to post
Share on other sites
The question, Nabla, is if the above Swiss based system, after looking at other possible methods, is in fact your ONLY choice to rank players effectively from #1-#72 when not all players have met all in combat.

Almost definitely not.

For we then don't have to invent the wheel, as Swiss play and their various tie-braking systems are very well documented.

Fortunately we would not be out of work. Chess is a (much more) balanced game, our outcome possibilities would probably not be in the set {0, 1/2, 1}, and the way adaptive scheduling would be done would have an effect on the expected distribution of the outcomes of a scenario. Admitted, the Swiss bridge system on the Wikipedia page seems to be relatively close to what I was thinking; notice that there too, game outcomes are converted to common units called international match points; our equivalent would be the Nabla scores of individual scenarios. The two best players of the first round play against each other, etc.

But why I am focusing on this when we have just decided that it's not feasible. I will take a look at the article you sent me. Unfortunately I have to work this weekend; I mean, real work, involving monetary gain. :(

Share this post


Link to post
Share on other sites
Timeframe of RoW VI as follows:

(1) Nablah reasoning of a tournament, which is a pair <N,A> where N is a finite set of at least two elements, which we will call players, and A is a zero-one matrix, such that its diagonal entries are zeroes and for all i, j ε N, i ≠ j, aij +aji = 1. The interpretation is that each pair of players in N played each other once, and aij = 1 if, and only if, i beat j.

(2) Nablah gets a bright spark, drops the "h" in his name, and produces another beautiful math equation for ranking/rating RoW in different sections. Failing that, repeat step (1);

(3) He writes a new front-end for NABLA, to be used by moi, the tourney director;

(4) Demo and CMBN by now released in wild. Everyone forgets NABLA and RoW for awhile;

(5) Scenario designers approached with concept paper for designing 7 scenarios specifically for RoW to test players abilities w/ every CM unit capable of being clicked w/ mouse pointer;

(6) Invited Scenario designers goes in evil genius lock-down mode and design diabolical battles, never seen before; the need to wear clothes while doing so not obligatory;

(7) Invites and dates set and announced on forum for tourney start. Vets that showed interest to play RoW VI will get 10-14 days grace to reply to invites, following that invites send to newbies, who showed interest in playing, all via PM.

(7a) (7) above not possible for Newbies without obligatory thesis titled: Why we should character assassinate your name upon drop-out;

(8) Great fun and gnashing of teeth to be had for +/-150 days, depending on state of intoxication while playing;

(9) Winners announced, I drink wine, send corks as prizes, winners fêted as the greatest ever to set foot on Normandy.

Thanks for the (very detailed) info ;).

Share this post


Link to post
Share on other sites

I finished the first and I *think* the second ROW tourneys, my nick at the time was Thumpre. Mind you, I'm pretty sure I finished LAST in those tourneys, but I'd like to have another go.....

Share this post


Link to post
Share on other sites
Hmmm... I wonder if my new scenario in progress, "For a few dollars more..." is evil enough..... might need to tweak it a bit....

:)

Keep the scenario a 'secret' for now, if you want to use it in RoW. We surely need your skills in designing some diabolical battles as per previous incarnations. Welcome back Spoon!

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Sign in to follow this  

×
×
  • Create New...