PDL 2021 "Alt-Scoring" Running Feedback Request

A PlayDiplomacy competitive league

Moderator: mjparrett

Re: PDL 2021 "Alt-Scoring" Running Feedback Request

Postby jay65536 » 01 Apr 2021, 19:43

OK, so in reviewing the thread, I think there are two major areas of feedback and a few other questions and notes. I'll start with the big ones and move down.

The two major areas are really sort-of one area with two variations: not enough separation in general, and not enough separation at the bottom.

Not Enough Separation In General--Too Many Tiebreaks, Not Enough Point Separation

mightyduk pointed this out, and while I don't agree with some of the arguments (I'll get to that below), the overall point is a good one. A system isn't very good if it doesn't separate the players on points and has to rely on tiebreakers.

I thought about this a lot when designing my system--would there be a lot of ties that need tiebreakers to resolve? Or would it sort itself out over multiple rounds to the point that the tiebreakers are a minor issue? That's actually a big reason I'm re-scoring the PDL games--to test this out.

Last year, when I re-scored the games, I found that tiebreakers were a minor issue. After 3 rounds, there were three pairs of players that were tied. One of those pairs was genuinely tied, and the other two pairs were separated by the center count tiebreak. After 4 rounds, one of the ties that used the tiebreaks was broken by different point results. So we only ended up with one case where the tiebreakers were used.

I don't think that's too much. But then again, last year could have been unusual. I will definitely be keeping an eye out for that this year.

As far as the complexity of the tiebreakers, which was the other piece of mightyduk's feedback, no one has asked what they are, but they are not complex. They're actually extremely easy to apply. For reference, I'll list the PDL tiebreakers (yes, they exist) first:

1. Most solos
2-7. Most 2way draws, 3way draws, etc.
8. Most centers in any one game
9. Most total centers (mathematically I think this one might be redundant, but it's in the rules)
10. Random draw

Here are the tiebreakers in my system:

1. Most solos
2. Most draws having outscored your opponents
3. Most overall board tops in draws
4. Highest score from any one game
5. Largest center lead in any draw where you outscored your opponents
6. Most total centers across all draws

I do not really think either of these sets of tiebreakers is very complex. The bit about how much they end up being used is definitely well-taken, though.

Not Enough Separation At The Bottom--Losses Should Earn Points

So, on a personal level, I do not like this idea, but so many people are in favor of it that I have been thinking about ways to incorporate this into my system.

First let me explain why I don't like it. When you vote yourself out of a draw, you are basically saying, "I concede that if the game continued, I would be eliminated." It is a useful tool to save time if everyone agrees what the result would have been, but to me it is just that--a time-saver, and nothing else. If someone would have gotten 0 points had they been eliminated, they should not get positive points because they gave up faster. The same logic can apply when someone loses to a solo.

That's why it's really confusing to me to see people saying things like, "We should reward losing powers for making it to the end of the game." In a draw-based scoring system, making it to the end of the game means making the draw. Voting yourself out isn't playing to the end; it's giving up early. Like I said, sometimes there's a time and place to give up early, but (in my opinion) not because you're going to be rewarded in the tournament structure for doing so.

Another thing that is avoided when you give all losses 0 points is you stop people from, to borrow someone else's phrasing, "losing your way to the top"--or to borrow from Oxmeister, "playing for defeat." This segues into what MJP was saying:

I don't know how much it really matters. League points may be different from system to system, but (the odd outlier aside) league placings won't be. I think if I remember correctly Jay published what D1 would have looked like last year under your method? About 90% of people were in the same place in the league with just a few shifting up and down one or two spots. The overall winner (and loser/relegation) places would have been the same under either system.


The bolded is not quite accurate, and I also want to point out something especially relevant to the last sentence.

I did indeed publish the "alt-standings" last year; anyone interested can look them up (it has its own thread). It is true that most of the time, if you took any two players, my standings would have the same one finishing ahead of the other that the regular standings did. However, I think it is noteworthy to look at not just how many differences there are, but where and how big the differences are.

In terms of where they are, some of them were in the middle or bottom of the standings. The ones that were, though, were noteworthy. For example, one player jumps from 18th in the standings to 13th in the "alt-standings". Another drops from 11th to 17th, and someone else drops from 14th to 18th. These are sizeable movements in the standings, even though they are only 3 players out of 21. If we care who finishes at the bottom for relegation purposes, that seems like potentially a big deal.

The other issue is where the movement is. Some of it was near the top! For example, the "alt-standings" switched 7th place and 6th place with the actual standings. That's not a big deal--but in a larger tournament with a top board, and if it were 7th and 8th instead of 6th and 7th, this would affect who gets on a top board. Finally, the player who was 4th in the actual standings moved up to 3rd in the "alt-standings." I don't know whether that already sounds like a big deal, but if the top 2 players had had their solos stopped, it would have determined the winner!

Many, but not all, of these differences come from the simple steps of giving all solos the same amount of points, and all losses 0 points. In particular, the 4th place/3rd place discrepancy is caused by one player having 5 more centers (20 more points in the PDL system) just in wins and losses than the other.

So to me it seems like it is important to not give any scoring incentive to people who do things like say, "I know I'm going to lose, just help me get some centers before I vote myself out" or "I know I'm going down, just help me make sure this other player gets a lower score than I do." I've experienced this a lot, and while I consider spite to be part of the game sometimes, it doesn't need to be rewarded by the tournament structure.

With all of that in mind--I have been thinking about a way to give a nominal differentiation to losing results, based on how many people seem to think it's a good idea. Here is an outline of something I think would be compatible with my system:

1. When I say "nominal," I mean something on the order of tenths of a point (except a solo might get 2 points in a non-zero-sum fashion--so this would break zero-sum).
2. It would not be center-based, but rank-based, mainly on order of elimination. In other words, the first one out would get 0 points for 7th, the next one out would get 0.1 points for 6th, etc., up to 0.6 for the top rank. Tied ranks would average the points (like Carnage).
3. If you make it to the end of the game and lose, you are considered tied in rank with all other players who did so--center count is irrelevant. Similarly, if you finish in the draw, you are considered tied in rank with all players who earn the same draw score as you--again, center count is irrelevant.
4. There would now be a new tiebreaker, and it would be the second tiebreak: fewest points from losses. (The trade-off would be that we should now expect the tiebreaks to very seldom be used, even compared to before.)

To the people who said losses should have separation, what do you think of this kind of idea?
jay65536
 
Posts: 590
Joined: 10 Sep 2016, 18:13
Class: Ambassador
Standard rating: (1120)
All-game rating: (1126)
Timezone: GMT-5

Re: PDL 2021 "Alt-Scoring" Running Feedback Request

Postby jay65536 » 09 Apr 2021, 21:25

D2 Jan G3

Result: Solo! 18/14/2.

PDL system scores it 324/56/8.

My system copies many others in existence (but obviously not all), in that solos are winner-take-all. So the winner would just get the full 180 points.

If the system were using the modification I proposed above, then instead of 180 for the winner and 0 for everyone else, the winner would get 182, the two survivors would each get 0.45 points, and the players eliminated would get 0.3, 0.2, 0.1, and 0 depending on who made it the longest before being eliminated.

I am guessing, based on prior feedback, that those who've been contributing in the thread are not fans of winner-take-all for solos?
jay65536
 
Posts: 590
Joined: 10 Sep 2016, 18:13
Class: Ambassador
Standard rating: (1120)
All-game rating: (1126)
Timezone: GMT-5

Re: PDL 2021 "Alt-Scoring" Running Feedback Request

Postby jay65536 » 28 Apr 2021, 17:46

D2 Mar G1

Result: 3way draw, 9/9/8/4/4.

The official system scores it 120/120/116/16/16.

In my system, this is another case of a rule that got applied to a different game: a 3way draw where no one got bigger than 12 centers. So the three draw participants would have split the points, 60/60/60.

I was following this game and I noticed (maybe it was stated in the AAR?) that there were a lot of endgame negotiations about who gets how many centers to make everyone happy with the draw. The two powers on 4 centers each negotiated with the Western Triple to get 4 centers each and then vote themselves out; the Western Triple had to decide who would take one fewer center (and thus 4 fewer points).

This to me feels like one of the kinds of games I designed my system for: three powers rolling the board all approximately the same size. In my system, there is no need to worry about the exact size of any power in this setup: it's going to be 60/60/60 and that's it. There's no quibbling for centers, and there's no penalty (other than one or two tiebreakers) for taking the draw on a slightly smaller center count than someone else. The only time that matters (in a 3way draw, at least) is when one power gets 13 or more centers. Also, there is no motivation to negotiate over how you are going to lose the game: once you decide that you will not be in the draw, you have no reason to hold up the endgame over a few points.

Do you guys think I'm off base in thinking that's a good thing?
jay65536
 
Posts: 590
Joined: 10 Sep 2016, 18:13
Class: Ambassador
Standard rating: (1120)
All-game rating: (1126)
Timezone: GMT-5

Re: PDL 2021 "Alt-Scoring" Running Feedback Request

Postby Pete the Great » 30 Apr 2021, 11:56

From my prospective we could have voted the draw earlier. I can not speak for those on the 0 side of the balance sheet. They may have entrenched in the corner to lengthen the game.
User avatar
Pete the Great
Premium Member
 
Posts: 110
Joined: 10 Feb 2019, 03:00
Location: NY State, USA
Class: Star Ambassador
Standard rating: 1151
All-game rating: 1162
Timezone: GMT-5

Re: PDL 2021 "Alt-Scoring" Running Feedback Request

Postby jay65536 » 01 May 2021, 19:28

Pete the Great wrote:From my prospective we could have voted the draw earlier. I can not speak for those on the 0 side of the balance sheet. They may have entrenched in the corner to lengthen the game.


Are you talking about the actual game under the PDL system, or how you think things might have changed under my system?
jay65536
 
Posts: 590
Joined: 10 Sep 2016, 18:13
Class: Ambassador
Standard rating: (1120)
All-game rating: (1126)
Timezone: GMT-5

Re: PDL 2021 "Alt-Scoring" Running Feedback Request

Postby Pete the Great » 02 May 2021, 00:17

How things might have changed under your system. The actual game is concluded, the draw proposal happened and passed for the reasons it did. (Italy and Turkey wanted some points, the winners were content with the scoring results) So under your system the game might have extended with IT entrenching in the corner if they did not want to vote themselves out.
User avatar
Pete the Great
Premium Member
 
Posts: 110
Joined: 10 Feb 2019, 03:00
Location: NY State, USA
Class: Star Ambassador
Standard rating: 1151
All-game rating: 1162
Timezone: GMT-5

Re: PDL 2021 "Alt-Scoring" Running Feedback Request

Postby jay65536 » 14 May 2021, 14:54

Was about to post another game but also just saw I never responded to Pete. Thanks Pete, this was exactly the kind of feedback I was looking for--others' ideas on how the gaming of the system might have changed. Very interested to hear if anyone has similar thoughts on the game I'm about to talk about below.

D1 Mar G2

Result: 6way draw, 7/6/6/6/5/4.

Actual system scores it 70/66/66/66/62/58.

Here's what happens in my system--if you never saw the 2020 thread, this may look weird at first, but it's not so bad. First let's say that another (the 7-center Italy) had topped the board by 2 centers instead of 1 (like, say, if it was 7/5/5/5/5/5/2 and the bottom power voted themself out). Then he would get 54 points, and the remaining 126 would be split 5 ways, for 25.2 points each to the others. However, because there were a total of four players within one center of the board top, those 4 players wind up splitting 54 plus three of the 25.2's, meaning they each get 32.4 points, with the 5- and 4-center powers still earning 25.2. The 7-center board top is basically the same as the three 6-center results except in tiebreaks (the board top counts for the 3rd tiebreak, and the extra center counts for the 6th).

So the final breakdown, after the math, is 32.4/32.4/32.4/32.4/25.2/25.2--basically a big "cluster" where some of the players earn a bit more than others but no one has a good score. In a general sense, that's the same thing that happens in the PDL system.

So any feedback on the actual scoring is of course welcome, but there's also a big "what-if" here that if anyone has thoughts on, I'd love to hear as well. In the PDL system, I think a big part of the reason the game ended here was because even though no one had a good score yet, no one really saw a path to a good score. There were two triple alliances on the stalemate line, and no one wanted to whittle anyone--which, in that gridlock, was the only way to move the scores in a meaningful way.

My system has 2 ways to shake up the scoring, and one of them could have applied here. Let's say that another had taken one more center, off of anyone (let's say one of the 6-center powers for argument's sake). Now it would be 8/6/6/5/5/4. With 8 centers and a lead of more than 1, another would now have gotten 57 points, leaving the remaining 123 to be shared 5 ways, for 24.6 apiece. That's an almost 25-point gain just for one more center! But more importantly than the number of points gained, 57 points is quite close to the 60 points that the equally shared 3ways are getting, which means that that 8-center board top is a perfectly decent way to keep pace with all the smaller draws. And it was only one center away--much more within reach than the alternate way to gain points, which is to eliminate someone.

Could the prospect of Italy trying this have shaken up the game dynamic? Would it have made the game more interesting and fun? And on a secondary note, does anyone reading this think it's a good thing that such an 8-center board top would have put him almost at the same level as an equally shared 3way? (Personally, I do.)
jay65536
 
Posts: 590
Joined: 10 Sep 2016, 18:13
Class: Ambassador
Standard rating: (1120)
All-game rating: (1126)
Timezone: GMT-5

Previous

Return to PlayDip Diplomacy League

Who is online

Users browsing this forum: No registered users and 1 guest