.net
All site revenue goes to charity

Subject: Should winnable variants affect Elo?


Date: Sat Oct 29 02:16:09 2022
User: Gulbis
Message:
Now that all the unwinnable games are rated properly, I'd like to start a discussion about the winnable variants. As it currently stands, you get the same points for winning a 4x10 game as you get for winning a 4x10w game. I don't think that's how it should be because winning the same game is more difficult in 4x10 than in 4x10w (because you don't know that it's winnable).

So my suggestion is to completely remove Elo points from the winnable variants. What do you think?

Date: Sat Oct 29 02:43:13 2022
User: joeygray
Message:
As previously stated, I disagree with that premise. The same game is in fact exactly as hard or easy to win regardless of whether you know it’s winnable. Your initial assessment, your plan, and your “mistake rate” (Hop, I know you don’t have one of these 🙂) are the same. Don’t see why the ELO result shouldn’t also be the same.

Date: Sat Oct 29 02:57:46 2022
User: Gulbis
Message:
I'm talking about cases where the "initial assessment" is that the game is not winnable (even though it is actually winnable). You can't tell me you would make the same assessment if you got the same game in a winnable variant.

Date: Sat Oct 29 03:01:01 2022
User: HopDiriDiriDattiriDittiriDom
Message:
"So my suggestion is to completely remove Elo points from the winnable variants. What do you think?"

No. Why?
Maybe a small adjustment with K factor (say 6 or 7 instead of 8) for winnables can be considered. Or leave it as is.


Date: Sat Oct 29 03:29:24 2022
User: joeygray
Message:
Yes, I can tell you that. I take every game as winnable, until I prove to the best of my ability it isn’t. If my ability falls short of being able to see all the way to the solution, as it surely does at times, that is what ELO is supposed to be measuring and how it should be operating. I will win a certain number of those by “if it’s winnable, then it’s got to be this way”, and the percentage of times I get that kind of thing right is also a measurement of my ability, to be captured in my ELO, AND the game’s ELO. Point is, I would make the exact same plays in these cases whether or not I knew the game was winnable. The game would not be easier, or harder.

If knowing the game is winnable forces you to look longer and deeper than you would have otherwise, that is also an aspect of freecell ability that ELO is to measure.

Date: Sat Oct 29 03:31:49 2022
User: TNmountainman
Message:
I of course agree with Gulbis - but of course everybody here knew that already - as previously stated umpteen times or more.

But Hop's idea is interesting. But I'd say keep the potential downside at 8, but the upside maybe at 4 or so. Possibly 5. After all...........knowing it's winnable ipso facto makes it easier to win, even if it's only 'psychological', shall we say. It's a huge step towards solving.

But really............Gulbis's last sentence is the 'purest' 'solution'. 😊

Date: Sat Oct 29 03:41:55 2022
User: cellmate
Message:

Elo is about winning or losing.

Players perceptions or strategies have no effect on the difficulty of the game.

If you change your strategy and lose, its your fault, not the variant's


Date: Sat Oct 29 03:42:34 2022
User: TNmountainman
Message:
While I *do* appreciate joey's viewpoint - I think it's too idealistic(?) to think most players behave that way. (And that's coming from a very strong idealist!) After all...........that's why some (most?) elo chasers opt for the winnables when available. I just think it's (nearly) undeniable that that's the motivation (or part of it).

Date: Sat Oct 29 04:26:34 2022
User: Gulbis
Message:
"No. Why?" (Hop)

Because of what I said in the first post. Difficult winnable games are more likely to get solved if you know they are winnable and that distorts Elo. Sure, reducing the points would help but why do that when we can just eliminate the points?

My question is - why not remove the points for winnable variants? Are there any downsides to doing it? I don't think so.

Date: Sat Oct 29 04:35:58 2022
User: sebcbien2
Message:
When I'm playing 12x0, I just play them as if they are all winnable games, but this state of mind may be not applicable for variants with a lot unwinnable games

Date: Sat Oct 29 04:56:44 2022
User: HopDiriDiriDattiriDittiriDom
Message:
"My question is - why not remove the points for winnable variants? Are there any downsides to doing it? I don't think so."

Your question sounds like to me "Why we should rate Magnus Carlsen's game (he's rated above 2700) when he's playing with a player rated below 2000?".

In practice, super GM's like Carlsen avoid playing with much lower rated players than themselves.

You can avoid playing winnables if you have some Elo-wise concerns of yourselves.

Date: Sat Oct 29 05:28:33 2022
User: Gulbis
Message:
I don't think your comparison makes any sense. I'm not trying to avoid low-ranked games or saying that they should not be rated.

My Elo concerns are for everyone, not just myself. Isn't it best if we have a fair system? Or at least as fair as possible? Removing the winnables would make Elo more fair, that is the clear upside. As for downsides, I don't see any.

Date: Sat Oct 29 05:38:44 2022
User: HopDiriDiriDattiriDittiriDom
Message:
"My Elo concerns are for everyone, not just myself. "

I am not sure what your concerns are. Are you worried about losing Elo points to a winnable variant game? Or you do not want gaining Elo points by playing winnable variants? Or what?

Your suggestion doesn't makes sense to me either. I do not think that knowing a game is winnable makes Elo scoring unfair.

The game is there whether you know it is winnable or not. You still have to win it. It could be an easy game or a hard one. You don't know it.

In my opinion winning a hard game (winnable variant or not) should be rewarded.

Date: Sat Oct 29 06:49:56 2022
User: strassenbeuge
Message:
When I'm dealt a tough game and I know it's winnable, I invest as much time as I can until I find the winning way. Of course, I make a careless mistake now and then and loose a winnable game. But all in all, I win more winnable games that I know are winnable than winnable games I don't know are winnable. That, by the way, is the reason I rarely play the winnable variants - it lacks that tingling uncertainty of whether the game is actually unwinnable or whether my judgment is wrong. Conclusion: If I would play only winnable variants, my ELO would be higher.

But what Denny always says: It's just a game. If there are minor injustices, I can live with that just fine. But I can also live well with it if they are corrected :-)

Date: Sat Oct 29 09:49:42 2022
User: TitanicTony
Message:
I completely agree with Hop:
"In my opinion winning a hard game (winnable variant or not) should be rewarded".

And, there are still a lot of never played (and never won) winnable variant games.

The winnable variants have been popular for years. They were a BIG improvement to the site, imo! Why, Gulbis, do you want to suddenly make them less popular?

Date: Sat Oct 29 10:27:02 2022
User: Kumquat-of-Conciliation
Message:
Denny has said quite a few times that winnables play is a small fraction of the total. As I recall, cellmate's data demonstrated that. They are obviously (relatively) more popular among the elo chasers, as a multitude of discussions have shown. Because they're easier to win. Maybe not inherently (nod to joey) - but because a player is already gifted a step forward.

Date: Sat Oct 29 10:34:34 2022
User: Gulbis
Message:
"I am not sure what your concerns are. Are you worried about losing Elo points to a winnable variant game? Or you do not want gaining Elo points by playing winnable variants? Or what?"

My concern is that those players who mostly play the winnable variants have an inflated Elo. Why? See strassenbeuge's reply. That's exactly what is happening with most people, me included.

I think part of the problem is that you think of Elo as a reward but it's not a reward. Your Elo score just shows your current skill level relative to the other players. That's it. And this skill level will be measured more accurately in the standard variants, not in the winnable variants.


Tony, I don't care which variants are popular and which are not. Anyone can play whatever they want. And who can say this change would make the winnable variants less popular? On the contrary, perhaps people are longing for a streak variant which won't affect their Elo.

Anyway, this idea is not sudden. I've mentioned it once or twice here before.

Date: Sat Oct 29 11:27:02 2022
User: HopDiriDiriDattiriDittiriDom
Message:
Interesting Gulbis,

You mostly play winnable variants yourselves according to my Hop-style Elo stats. Hence your biggest gains come from winnable variants, in which I include quasi-winnable ones such as 7x5, 10x2.

x7 3 2 2.021 14.648 -12.628
4x8 2 0 0.529 0.000 0.529
4x9 3 0 0.370 0.000 0.370
4x10 1 0 0.089 0.000 0.089
5x6 6 1 1.366 7.778 -6.412
5x7 10 0 1.311 0.000 1.311
5x8 16 0 6.881 0.000 6.881
5x9 62 0 22.234 0.000 22.234
5x10 2 0 0.011 0.000 0.011
6x5 7 0 0.302 0.000 0.302
6x6 9 0 2.740 0.000 2.740
6x7 33 0 11.460 0.000 11.460
6x8 269 1 21.639 7.893 13.747
6x9 1 0 0.029 0.000 0.029
7x4 12 0 0.766 0.000 0.766
7x5 138 1 156.153 7.651 148.501
7x6 49 0 2.708 0.000 2.708
7x7 226 0 2.011 0.000 2.011
7x8 2 0 0.001 0.000 0.001
8x3 9 0 1.825 0.000 1.825
8x4 33 0 1.059 0.000 1.059
8x5 36 0 0.302 0.000 0.302
8x6 111 0 0.380 0.000 0.380
9x2 26 0 1.941 0.000 1.941
9x3 52 0 7.372 0.000 7.372
9x4 46 0 0.256 0.000 0.256
9x5 238 0 0.402 0.000 0.402
9x6 3 0 0.001 0.000 0.001
9x7 5 0 0.002 0.000 0.002
10x0 44 1 8.640 7.817 0.822
10x1 4 0 3.095 0.000 3.095
10x2 1200 0 99.430 0.000 99.430
10x3 40 0 0.127 0.000 0.127
10x4 111 0 1.071 0.000 1.071
10x5 13 0 0.014 0.000 0.014
10x6 41 0 0.010 0.000 0.010
11x0 11 0 0.858 0.000 0.858
11x1 14 0 0.660 0.000 0.660
11x2 113 0 0.213 0.000 0.213
11x3 172 0 0.214 0.000 0.214
11x5 1 0 0.000 0.000 0.000
12x0 33 0 1.850 0.000 1.850
12x1 79 0 0.229 0.000 0.229
12x2 145 0 0.049 0.000 0.049
12x3 4 0 0.000 0.000 0.000
13x0 52 0 0.756 0.000 0.756
13x1 378 0 0.212 0.000 0.212
13x2 35 0 0.009 0.000 0.009
13x3 3 0 0.002 0.000 0.002
4x10w 30 1 25.769 7.078 18.691
5x8w 46 1 33.624 7.708 25.916
6x5w 38 2 10.041 15.772 -5.731
7x4w 74 1 51.317 7.799 43.518
8x3w 56 0 12.747 0.000 12.747
9x2w 178 1 78.895 6.356 72.539
10x1w 1271 5 404.965 34.605 370.361
11x1w 274 0 10.570 0.000 10.570
12x1w 79 0 0.156 0.000 0.156
13x0w 221 0 4.209 0.000 4.209

Date: Sat Oct 29 12:30:11 2022
User: Gulbis
Message:
Yes, I know which variants I play :) But what's your point? That my Elo is inflated? Perhaps. Does that mean I shouldn't want my Elo to be accurate? Of course not. I don't want my Elo to be artificially high. I don't want anyone's Elo to be artificially high.

Date: Sat Oct 29 13:53:11 2022
User: redberet
Message:
It doesn't really matter why a player chooses winnable. That's their prerogative. Why should they be penalized? I hold no grudge on how anyone gets their points. It's all for fun.

Date: Sat Oct 29 16:28:04 2022
User: BuzzClik
Message:
"Does that mean I shouldn't want my Elo to be accurate?"

Accurate? It's an artificial metric based on an algorithm with mostly arbitrary parameters.

One of the most ironic things about Elo is that very early on someone apparently was losing a ton of games in a row to *somehow* manipulate their Elo. A stern warning was giving that manipulating the Elo would not be tolerated. BUT many Elo discussions center on how to inflate one's Elo.

One important question: can one exchange their Elo points for NFTs? Bitcoin? Coupons for Cap'n Crunch?

Date: Sat Oct 29 17:01:28 2022
User: Kumquat-of-Conciliation
Message:
It sure would be helpful if Denny would provide hot links for the current exchange rates. It's quite the hassle to have to keep all those other tabs open. Several times I've seen the rate for bitcoin and dogecoin vary as much as 10% *during* a tough 9x2.

Date: Sat Oct 29 17:22:33 2022
User: Kumquat-of-Conciliation
Message:
I also wonder how many of the pics posted here over the years have been turned into NFTs. Probably a lot, as there's some good stuff therein.

Date: Sat Oct 29 20:57:58 2022
User: roo
Message:
"inflate one's Elo"

Very clever BuzzClik.

To add to that, I have not seen any mention of associated error bands for the Elo number. They would be significant enough to render this whole discussion pointless I should imagine.

Date: Sat Oct 29 21:23:34 2022
User: redberet
Message:
That would be ElGuapo's domain to explain. That's what we sent him to school for.

Date: Sun Oct 30 02:30:02 2022
User: Gulbis
Message:
redberet, players would not be penalized for playing the winnable variants. They would neither get nor lose any points. It would just be neutral.

As I said before, your Elo score is just your current skill level. I'd like it to be as accurate as possible and excluding the winnable variants would help with that.


As for how much it really affects, my guess is that the winnable variants affect the score very little for players outside the top 20 because they lose enough games anyway and it all mostly evens out. It's the top 20 players that most likely have an inflated Elo from playing winnables, and especially the top 10.

Date: Sun Oct 30 03:11:40 2022
User: redberet
Message:
I took a look at the recent play of the top 20 and there is only one who sticks solely with winnables. Even given that, who's to say that the others are adhering to any particular set of ethics? No one knows and that's the point.
Taking care of a perceived problem can never have satisfying results.

Date: Sun Oct 30 03:37:17 2022
User: Gulbis
Message:
This has nothing to do with ethics and this is not a "perceived" problem. It's a real problem. Minor, to be sure, but still real. strassenbeuge explains it very well.

The fact that no one so far has been able to come up with any downsides to excluding winnables from Elo is also telling.

Date: Sun Oct 30 04:58:37 2022
User: TitanicTony
Message:
Not true. Hop put it very well, imo: "In my opinion winning a hard game (winnable variant or not) should be rewarded".

So far, I've not seen a single "vote" in favour of your suggestion, so I think it is not going to happen.

Date: Sun Oct 30 05:30:35 2022
User: Gulbis
Message:
What is not true? Is there a downside? I haven't seen it. "No rewards" is not a downside because ELO IS NOT SUPPOSED TO BE A REWARD. I don't know how much more clearly I can say it.

Elo is supposed to be a measurement of your skill. The more accurate the Elo score is, the better. Do you not agree?

What is the point of Elo? It's to rank players, isn't it? So that we could say player A is better than player B, right? And isn't it true that the rankings will be better if the score is more accurate?

-----------
I don't get this stuff about rewards. I think people play Freecell because they enjoy it and because it stimulates the brain. Solving a hard game is its own reward, it gives you a feeling of satisfaction. We all played Freecell without any "rewards" before Elo was introduced. And now all of a sudden we need rewards? Well, you can get yourself a piece of candy or something as a reward if you want. There's no need to use Elo for that.

Date: Sun Oct 30 05:47:13 2022
User: HopDiriDiriDattiriDittiriDom
Message:
Of course I mean by "reward" here, in a loose sense, "some rating points added to your Elo score".

Gulbis seems to take it more seriously than that.

Date: Sun Oct 30 06:07:39 2022
User: sebcbien2
Message:
may be it's beyond winnable variant vs streak variant, as they could be tough winnable games
isn't it to count all games vs count only hard games ? it's what Elo system is supposed to do

player A is able to solve 300 easy games a day without loosing a single one
player B can solve one very very difficult game but only one per day
who's the best player ?
we might have like pokemon card game, a multi criteria measurement: stamina, strength, hability, speed...
so we can compare ourselves in the same skill area

what about an average Elo points gained per game

Date: Sun Oct 30 06:35:32 2022
User: Gulbis
Message:
Hop, I would like Elo to serve its original purpose, which is to rank players. If you think that's too "serious", well, I don't know what to tell you. Sure, we can all decide that Elo's main purpose on this site is to be a gimmick and to give out rewards. Is that really what you all want? Sounds like maybe yes.

I must admit I'm disappointed because I enjoy the competitive aspect of Elo and would enjoy it even more if the scores were more accurate.

Date: Sun Oct 30 06:49:32 2022
User: redberet
Message:
No player is rewarded points for a win, they are awarded.

Date: Sun Oct 30 07:21:03 2022
User: TitanicTony
Message:
@Gulbis. You will almost certainly remain in 2nd place, no matter how the Elo is scored, 🙂!

Date: Sun Oct 30 08:00:19 2022
User: HopDiriDiriDattiriDittiriDom
Message:
"I must admit I'm disappointed because I enjoy the competitive aspect of Elo and would enjoy it even more if the scores were more accurate."

That can only be achieved with a complete reset.
Earlier I stated that (in a different thread) I'm ready for it. But I don't think Denny will do it. My Hop-style ratings are closer to what you want (maybe) to get a better idea of it. There is a smaller margin between you and I for example.

But then again not rating winnable variants doesn't make sense to me. There is big subset of games shared by the winnables and regular ones
and you suggest playing them as winnables should not taken into account because they are known to be winnable in that case. That's odd.

8x4 is winnable variant also and we play it knowing beforehand that all the 8x4 games are winnable. They should be excluded by your rationale as well.

Date: Sun Oct 30 08:21:19 2022
User: Kumquat-of-Conciliation
Message:
Not trying to speak for Gulbis, but I think his/our point is cases where there are both options. There exists for 10-1 game xxxx both a version that's known to be winnable, and one that's unknown. Our 'claim' is that the known winnable is easier to beat. The scoring for 8x4s, for example, would not be affected.

Date: Sun Oct 30 08:41:52 2022
User: HopDiriDiriDattiriDittiriDom
Message:
But it boils down to "They should not be rated bacause they are known beforehand that they are winnable".

This can't be the reason not to rate them IMHO.

Date: Sun Oct 30 09:47:51 2022
User: ElGuapo
Message:
Hey gang, sorry I'm late to this. Wasn't feeling well. Ratings discussions are frustrating because it's not the most intuitive thing and people's understandings differ. I know Gulbis gets what we're aiming for and I appreciate that.

We're rating winnable play mostly because we can and the more data the better. HotStreak and tournaments are timed events with different incentives, so all that play is wasted in terms of ratings. Winnable play though is identical to regular streaking in every respect but one: the potential distortion caused by players knowing the game is winnable.

I don't like that any more than Gulbis does, but we probably disagree a bit on the magnitude. Remember this affect only exists at all if both modes of play are happening in a particular game level. In other words, if all the 10x1-10 play is happening in winnable mode then there's no distortion. The model will simply rate that game level lower.

Of course real life is more muddled than that and there will always be someone playing winnables in a game level with mostly standard play. I'd love to see these two play modes merged back together eventually. Our community's play is way too fractured already. And I fear that making winnables the non-rated way to streak just furthers that division.

Date: Sun Oct 30 10:06:14 2022
User: Gulbis
Message:
"But then again not rating winnable variants doesn't make sense to me. There is big subset of games shared by the winnables and regular ones and you suggest playing them as winnables should not taken into account because they are known to be winnable in that case. That's odd."

Why would that be odd? The standard variants are a complete set, they cover all the games. It's more than enough for Elo. But the winnable variants are different (easier) so they distort the ratings. Don't you really see it? If you get a difficult game (I'm talking about people in general here) where you don't see any path to a win, or even to clear a column, you might give up after 5 minutes or so. But you are more likely to spend more time on it if you know for a fact that it is winnable. So this game might not get solved in the standard variant but might get solved in the winnable variant. This means your Elo score is now 8 points higher than it should be.

Of course, there are exceptions. Joey says he doesn't do that, maybe someone else as well. But most people do play like that because it's human nature.

Date: Sun Oct 30 10:16:20 2022
User: MrFixit
Message:
Let's let ElGuapo weigh in. But my opinion:

1. it's just a different way of playing the same games.

2. There's probably some small delta in win rate based on info that you know in advance it's winnable but it's negligible compared to other issues. There are plenty of ways to game the Elo system, one of which I'm doing by playing the variants with the highest Elo. They are slowly adjusting. The "payout" on 7x6-12 went from about 0.5/game down to about 0.1/game in the relatively short time I was playing those. But ain't no way I'm really a ~2000-level player. I'm just exploiting something in the system for lesser played variants that's slowly adjusting itself.

3. We've gathered data this way for years.

4. And my usual cry in the sky, "it's the same for everybody". If you can mine the microscopic delta, go for it.

Elo isn't everything, it's not "perfect" and never will be, it's just yet another a yardstick.

Date: Sun Oct 30 10:21:18 2022
User: HopDiriDiriDattiriDittiriDom
Message:
"So this game might not get solved in the standard variant but might get solved in the winnable variant. This means your Elo score is now 8 points higher than it should be."

If you really believe Elo scores differs that much, just by playing a winnable variant, that's odd (just my opinion).

Your argument is not different from saying that "if a game know to winnable, it should not be included in the Elo calculation". All the 8x4 games are known to be winnable. So they should not be included either by your rationale.

Date: Sun Oct 30 10:38:59 2022
User: Kumquat-of-Conciliation
Message:
No, Hop, that's not the same. Only when there's a 'choice' between winnable or not.

As to the 8-pt. difference in elo score, it's like strassenbuege wrote above...................(for most players not named Hop or joey)...............they'll spend more time on trying to win a game known to be winnable. So you might eventually solve a difficult game, maybe adding, say, 1 elo point, VS...............if you had given up, yet it turned out to be winnable, that's a loss of ~7 or so points -- thus the 8-pt. difference.

IMO, it's a larger delta than MrFixit thinks. Yet still...........as he notes, there are other ways to game the system, unfortunately. But it's a whole lot better than it was 6-8 months ago.

Date: Sun Oct 30 10:43:11 2022
User: cellmate
Message:

i cringe whenever someone claims that knowing the game is winnable makes it easier

Games don't change. Players do.

The premise that the game is easier because its known to be winnable is bogus because games don't change.

The premise that players win more often when they try harder is fact.

Perhaps knowing the game is winnable is an advantage for some. If it makes them try harder, all the better. Players win when they try harder.

No reason to discount Elo when you try harder.


Date: Sun Oct 30 11:03:39 2022
User: redberet
Message:
I'm curious if not for points, why do you play winnables, Gulbis?

Date: Sun Oct 30 11:15:50 2022
User: MrFixit
Message:
Winnables

Cols Cells PctWinnable
5 10 96.2234
6 8 99.2584
6 5 96.5179
7 4 99.3449
8 3 99.6343
9 2 99.2727
10 1 95.9696
11 1 99.6394
12 1 99.9797
13 0 96.7967


Date: Sun Oct 30 12:37:16 2022
User: Gulbis
Message:
"I'm curious if not for points, why do you play winnables, Gulbis?"

Why, for the streaks, of course. If you try streaking in the standard 9x2 variant, for example, sooner or later you'll stumble upon an unwinnable and your streak will be snapped even though you did nothing wrong. I appreciate the winnables for this reason - whenever you lose it's your own fault.

Date: Sun Oct 30 12:50:08 2022
User: Gulbis
Message:
"Your argument is not different from saying that "if a game know to winnable, it should not be included in the Elo calculation". All the 8x4 games are known to be winnable. So they should not be included either by your rationale."

I'm sorry but that's preposterous. It's not at all the same. Kumquat already explained it to you. In my example there were two possible situations - the winnable variant or the standard variant. 8 point difference. That's when Elo gets distorted. For 8x4, there's just one possible situation and that doesn't cause any Elo distortions.

Date: Sun Oct 30 13:05:10 2022
User: TitanicTony
Message:
I agree with cellmate! This whole thread makes me "cringe"!
I really wish this thread had never been started.
ElGuapo makes a good point. It is not just the players that are being "rated", but also the games. Eliminating the winnable variants would be throwing out a lot of data! That is definitely a "down side", imo!

Date: Sun Oct 30 15:29:55 2022
User: joeygray
Message:
Gulbis, your passioned argument about how good it is to make improvements is beside the point. It is not at all clear this would be any improvement. Obviously you think it is, but a number of folks don’t.


Post follow-up
Username: New user? Create a free account here
Password: Note: username and password are case-sensitive
Message:
Editor by summernote.org
Email notification:

All content copyright ©2024 Freecell.net
By using our games you consent to our minimal use of cookies to maintain basic state.
Maintained by Dennis Cronin