Board Game Arena

I once read some lines about this, but only as a side-topic in a discussion about general Elo if I remember it correctly.

Sometimes it can feel that your point differences in wins are larger 90% of the time than when you lose.
I have had streaks of four consecutive close 2nds in 6 nimmt by under 6 points maybe, followed by wins of more than 30. Or streaks where every Dragonheart I would lose was within 0-10 points. Same holds for Color Pop (<2 points in comparison to not seldomly >5 points in wins) or Stone Age.

So while if you take into account the sheer points collected over some specific period, you could arrive at an exact draw, while in a really bad spell the actual result could nevertheless look like 1-9 to your disadvantage. That's naturally an outlier, but against players of similar level anything but impossible. Namely if you consider the points as primary, it's indeed accurate that they are very close, but the distribution among games can be random/arbitrary. If could befall that your luck and ability concentrates on 1 of 10 games, so that your gap would be maybe +18, while in the others -1, -2, -4, -7... If you could manipulate the results, you would substract 17 points from your win and distribute it that way that you get maximum outcome from the remaining series, so you add +2 to the -1 = +1 = closest win, and so on.

Normally you can assess an average gap by the difference of skill in players involved, and the actual results will in theory describe a bell curve around that gap. The unfairness comes into play when it's expected that your (rare) losses will always contain very small margins, and a heavy number of your wins very high ones. That's for example when you take something with 600 vs. 150 or something, and in a game that mainly is based on skill, but can also consist of random elements that can lead to heavy imbalance (again Stone Age could be an example: how do the dice fall, who gets those early farms, tools by shop cards? Who sees the best cards first at average so the first move is a bigger advanatge? And particularly devastating can be, in whose round that mighty 1-7 hut pops up at the end, when many resources are involved?). This way, winning a lot of games by 80 or more points against someone while losing the odd by less 10 points doesn't sound unrealistic.

Besides the claimed fairness, another essential upside for me is that there is less frustration/hopelessness at early foreseeable losses. You would make an effort to reach the best possible result. The undifferentiated yes/no (win/loss) is omitted in favor of a continuum, just as the point tally represents a continuum (if you exclude the games that end with 1-0 in any case, mostly abstracts presumably).

So the consequence is that the Elo calculation would be more accurate and the matches more engaging (less perception of 'it's over anyway, when can I finally resign') even when they seem to be decided.

I would be very careful with any mechanism that encourages "playing to minimize losses" over "playing to try and win" for anyone who's behind. We already have this situation in free-for-all games with 3+ players, where you might be inclined to go for second place instead of last place due to Elo computation. Adding even more of these problematic incentives, especially to two-player games, could make BGA less enjoyable as a whole.

Also, the point I'm bringing up regularly for suggestions similar to this one: The scoring function for many games doesn't actually indicate how close the game was. Check Lords of Xidit for one example, Innovation for several others, as mentioned in https://boardgamearena.com/forum/viewto ... =9&t=13618

When I'm about to lose by a small margin, I usually take high-risk moves which with a low propability get me the win or with a high propability make me lose by a very big margin. And I think that's the right move to do in such a situation. The mechanism you propose however kinda turns it into a bad move.

Also if I see I'm gonna win big, I sometimes go easy on my opponent to make him not feel totally crushed, because I think that's the nice thing to do. I don't wanna trade playing nice to maximizing ELO.

I wish I could like Silene post. With (heart)

A win is a win.

Over a large period of time I can only see that 'small margin of loss' problem should even itself out amongst players.

If you have a strategy that wins say 20% of the time but when it does wins by BIG margins, then this then could become the optimum strategy to play to gain ELO points. BGA is supposed to try and recreate the home board-game experience, ELO gains are never in the rules.

AlphaGo was noted as playing a strategy with moves that increased the chance of winning over moves that could win by a greater margin.

Due to recent events, I'm feeling a bit urged to revive this topic a little.

First, I understand all the objections. Some of them I hadn't been aware of, for example that not in every game the exact result reflects the difference in quality. This is due to my exposure, the games in my profile largely or completely don't belong to that class. What Silene said: Being 'nice' and not humiliating the opponent is honorable, but essentially your style and your own choice. To me it appears a little blurred or generalized, because losing on its own (not depending on the margin) can feel sufficiently bad to a more ambitious or experienced player. On the part with high-risk moves that win rarely, the crucial point is, everyone would know about the new system and therefore can choose the now modified optimal play; within the scope of his ability. If my proposed system of gradual ELO calculation had been implemented here from the start, I could in return bring up the same point how artificial it would be to commit to such 'all-in' moves, rather than opting to max out your point score. It's a matter of habituation, as so often.

@RicardoRix
On a few occasions in other threads already I've got the impression that you don't actually read my lines in depth, or just interpret it in a way that suits you only to debunk this now false and thus void statement. I pointed out explicitly how the curve of winning margins would be between uneven players, and that the better who maybe wins 90% in total could only win 50% or even less of close games (<10 points or any other number according to valuation). From that we can easily conclude that it doesn't 'even out'. And as I painfully learned in some games, neither among more equal players level. Solely in theory the percentage of head or number in coin flips will approach more to 50% with rising number, but that's the relative difference, not absolute. Because it should be feasible if we consider n=1 throws, the difference will be 1, while with n=500 I could only make a guess with what probability it's not higher than 1 (necessarily meaning both head and number appeared 250 times), but it's pretty certainly lower than 2%. That's because probability has no memory, or whatever the saying is in English.
About the rest, same as with Silene, no big deal for people to adjust to a rule that overt for everyone. The problem just arises if that rule is innately unfair, which I don't see yet (or still). Apparently these additional features - ELO, statistics, replays - don't conform to the home experience as they are rightly believed to improve the overall experience. And so my request is only to create high fairness and consistency: If you already have a system that is meant to assess player skills, it should be perfectly accurate. What I said in the other thread about differentiated results in abandoned games: It's just not fair when with over 50% game progression, or even over 80%, when often the winners / top places are settled, they don't get rewarded because of a random event no one can influence (a quit or disconnect by someone). AlphaGo is no human and despite, I'm sure the professionals could program it so that it would include the winning margin in the assessment. Then again, Go only shows results of 1-0 here, so that would be kind of futile. Or it would support the assumption that there are games that are so interwoven that a subtle move early on will maintain your advantage that's just about enough to win eventually. But that's the main point here, to distinguish between this and games where more points automatically means better. You surely don't want to waste 10 points at any stage in Stone Age, or Dragonheart, or 6 Nimmt (these come to mind at first from my personal history).

All this is motivated by my experiences that my average loss would be around half of my average winning margin (in games with an actual points score of course, unlike the just mentioned Go for example), and that's still a cautious approximation. So at last the specification that lead to that posting, my recent Stone Age results. It just happens that I had unusually frequent 'close' wins for my standards, and still they nowhere near are in the same bracket in terms of relative numbers as the close losses.
https://boardgamearena.com/gamestats?pl ... game_id=13

Four games this evening, wins by +73 +63, losses by -2 -4.
From the last 8 losses 6 were within -15 points.
In the same span since the first loss (49 games), 10 wins are within +15 points (three games exactly that value though, so 7 with +14 or less). If I counted right, there were 21 wins with over +50 difference, the one biggest loss was -31.

Just to emphasize, my intention here is not to complain, but rather to possibly find a calculation method that actually leads to less discomfort. Because as mentioned in the opening post, when you happen to be on the adverse side it can be a little discouraging long term. And to pick up that one thought again, you can play equally good in two sequences of 10 selected games and still lose 100 points more ELO than in the other. Because if you do five 'mistakes' total (lets say a move that costs you 10 points in the result), and the majority in the really tight games (lets say four of these ten games), out of pressure or tiredness or else, you could in this made up scenario lose them by -2 -8 -5 -7 or whatever makes the mistake damaging. By contrast, if the mistakes occur in the other six rather comfortable wins (or at least 4 mistakes, which is not a far stretch), you save three losses and come off 60 ELO points better. With the same overall performance though, as described.
What I haven't brought up yet too is that not only do hopeless games become less 'hopeless' and thus reducing the feeling of wasting time the remaining match, the opposite applies with the victories. I occasionally sensed some tendency in the past to deliberately make moves I knew were not optimal just to avoid the 'boredom' emerging from the awareness of the game being decided in my favor. That would be eliminated too, so everyone has incentive to do their best, which would perhaps even reduce concentration lapses also in the closer games (because these lapses could spread when you are not used to play to your limit). And that should be the ideal attitude in my book no matter if you are more of a competitive or casual gamer.

Or.

OR

You could ignore your Elo, and just allow the site to calculate it however it wants, and just use it as a way to get better than terrible matchmaking.

Or.

You could practice not becoming offensive on a post that contained nothing of that attitude, but instead many objective facts and proposals.

In effect, I belong quite comfortably more to the casual gamer group, while only/mainly my affection for statistics might countervail that significantly. Sure, if you look at this Stone Age chart, you can get the impression I avoid the highest ranked players. But you have to consider several things.

1. While it is a well established game with many capable players, you will not encounter experts all the time.
2. I often play it as a game in between where I don't wish too much thinking/pressure/involvement, and still, most of the opponents are in the 350-450 range (indeed I avoid under 250 nowadays almost at all costs), which feels like an acceptable compromise for me.
3. And this above all. It's a long game and so putting much effort into it it can be more frustrating to have a streak of bad results. Again, the method of calculation only works detrimental in this regard, because if it were like I suggested, I am not so negatively affected by close losses, which would lead to less susceptibility to only a certain player level.

Edit.
btw here is the proof that led to a bad experience, just supporting my thesis that seeking much better players will only mean many close losses, but almost none over 40 points or so... qpenguin got the same reward for a game that was basically equal (just change any one move of that game against his favor perhaps and he will forfeit at least that crucial one point, even something as slight as rolling one ressource less for him in the last round) that others get for the occasional 100+ points win.
https://boardgamearena.com/table?table=52853108

There might be particular games where this would make sense but in general I'm against it and it probably wouldn't be worth it to figure out which games this would be good for.

If I know I can't win unless I try for a strategy with a low probability of success (attempting to top deck just the right card, for instance) but will on average tend to result in a lower score than different action with lower or no probability of winning, I should no be penalized for taking that chance at all in two player games, and not beyond the possibility of falling behind other players in games with more than two players.

If ELO rating would encourage winning by bigger margin, wouldn't it at the same time encourage selecting weaker opponents?

IMHO, close games are a lot of fun (they have more tension til the end), while the games with a clear runaway leader are a bit boring (for both winners and losers).

PS: Sorry if I didn't understand the suggestion.