Winnable variant = a kind of user-friendly one
Others = Not-so-user-friendly kinds (you may choose to play at your own risk)
>>Date: Sat Jan 7 07:02:33 2023
>>User: outskirts
>>Message: "Suzanne, eating the rice cake..."
That's a cute clip, but there are lots of people around the table, and none of them are being passive and sweet. A better example: An aging group of characters from the Batman stories is gathered at a table discussing diets:
MAYOR OF GOTHAM CITY: This new diet of rice cakes with a schmear of jelly is really rough
POISON IVY: Is it really a diet, or are you "cheating" <air quotes and a wink, wink>
BATMAN: Plain rice cakes is the only true diet. Everything else is a compromise of virtue
POISON IVY: Always the pompous jerk. I used to be on a diet, ages ago. It sucked. So I quit dieting. Why are you always forcing your "morals" on everyone?
JOKER: Here's an idea, Batman: use this gun for a nice game of Russian Roulette. See how that helps your diet!
POISON IVY: <laughs loudly> Or maybe start the ultimate diet: starve yourself to death. Maybe then you'd lose some weight, you fat pig. Or at least you would finally shut up.
BATMAN: Diets are meant to be difficult. It is all part of the transcendental experience.
TWO FACE: I am a strict adherent to the rice cake diet. I just finished one. See?
BATMAN: You admitted that you sneak back to your room and eat candy bars and corn chips. That's cheating.
TWO FACE: You show me a legal document where it says that I can't eat snacks between rice cakes!
BATMAN: Do you really need written rules as your moral compass?
POISON IVY: Quit harassing him, fatman!
BATMAN: Juvenile insults? Really?
POISON IVY:You hate me because I'm a woman
RIDDLER: Batman, you only criticize other people's diets because you are obsessed with the paleo diet.
BATMAN: When was I ever on a paleo diet?
RIDDLER: You are obsessed with it.
POISON IVY: That's right, fatman. And don't think that I'm flirting with you. I find you repulsive.
BATMAN: What is your problem?
POISON IVY: Quit bullying me!!!!!
@ElGuapo (mostly) ----
A probably earlier-answered question for winnables............. (Sorry, I moderately perused some other threads with "winnables" in the title but didn't come close to finding the answer. I suspect it was much earlier.)
Ok........so you all know I practically never play winnables unless it's maybe a WWC I take part in - and even rarely those. But tonight I was in the mood to play some 9x2, but was too tired to play the regular ones, so I went weird and tried some W9x2s. I know it's a limited sampling, but there seemed to be more elo points awarded than proper for the difficulty. But my question is not that - but it made me think..... It's whether or not the winnable-ness of winnables has been factored into elo points. (I also don't chase those, so this is a bit out of my lane.) I fear I know/suspect the answer is "no". But now that almost all (with some significant abstainers, I realize) have accepted that winnables are 'easier' (to win, specifically) than regulars, is it time to incorporate that 'data' into elo? I suspect the answer to that is also "no" for the reason that (supposedly) over time things will correct themselves. But can't/shouldn't that process be accelerated to reflect 'known' reality?
I don't expect to find any support for my idea, but I think it has merit. But mostly, I just want to confirm for sure that such was not considered in elo ratings. And of course this would also explain/confirm why the elo hunters play winnables so much.
I don't know how you can interpret it but this is from my "Hop style Elo" page:
I assume that's from *your* play? I *think* I understand what it means.
And you've played over 140,000 W9x2s?!?
Edit: Or is that representing *all* players?
Further edit: If the point is to show that the differential is very close, it doesn't take into account that (in general) weaker players would tend to play the winnables more.
All plays and yes I want to point out that the ratios are very close:
Edit:
"it doesn't take into account that (in general) weaker players would tend to play the winnables more"
I believe it is taken care of by Elo.
"But now that almost all (with some significant abstainers, I realize) have accepted that winnables are 'easier' "
TN logic
This was discussed at length in this thread.... Should winnable variants affect Elo?
(unfortunately the new new post editor lead to losing previous message formatting so its harder to read)
Some numbers for 2022...
Streak vs Winnable
9,2 Layout (all levels)
49452 Streak games played
35914 Winnable games played
14838 Games played in both
5581 Compared (games with same number of plays in both)
5 Max plays per game
4305 Games with same number of wins in both
899 Regular has More wins
377 Winnable has More wins
If winnable is removed from ELO you take away 8x4, which is probably the only variant that has accurate ELO's at the game level.
I'm not suggesting take elo away - just wondering if it shouldn't be tweaked for those variants where there are both flavors.
Will look at the posted data later when I have more time...
if I check my stats, I'm better with regular 13x0 than 13x0w but I'm not really representative
13x0
Win% = 96.66 %
Adj Win% = 96.65 %
Only Winnables Win% = 99.12 %
Win% per year : {"2021":{"%":"97.29","Total":295},"2022":{"%":"96.89","Total":5948},"2023":{"%":"96.31","Total":4368}}
13x0w
Win% = 99.08 %
Adj Win% = 99.06 %
Only Winnables Win% = 99.08 %
Win% per year : {"2021":{"%":"99.11","Total":1580},"2022":{"%":"98.94","Total":3391},"2023":{"%":"99.26","Total":2293}}
and same for (with much less played games)
9x2
Win% = 93.68 %
Adj Win% = 93.10 %
Only Winnables Win% = 93.68 %
Win% per year : {"2021":{"%":"93.75","Total":32},"2022":{"%":"93.98","Total":83},"2023":{"%":"93.22","Total":59}}
for 9x2w
Win% = 92.05 %
Adj Win% = 91.39 %
Only Winnables Win% = 92.05 %
Win% per year : {"2021":{"%":"93.94","Total":33},"2022":{"%":"92.65","Total":68},"2023":{"%":"90.00","Total":50}}
I can now see that why I couldn't locate the proper thread(s) was because I searched for "winnables" instead of "winnable". I think that's the second time I've done that. Sheesh....
@ Hop........... So there have been only 80,000+ games of regular 9x2 played, but over 140,000 of winnables?? That doesn't seem right at all. The winnable flavor is MUCH younger, so even if by chance they were played at the same rate since that introduction, there should still be many more regular plays than winnable plays. Seems to me.....
9x2 has over 4 million plays. Hops might be from his ELO tracking totals.
The winnable games are NOT "easier", they're the same game as the regular, but they've been proven to be winnable. They may SEEM easier to some, but they are the SAME game. Certain people want to label them as easier, but it's not true. I prefer playing winnable games because if I lose, it's because I messed up rather than the game was "unwinnable". It kinda reminds me of Russian Roulette.
The results here are based on the data of every game played each day starting 9/13/2021.
Hop, as usual, you are too far past me, programming-wise, for me to interpret what all you're saying. First of all..............is it correct to say that you have formulated your own elo calculations, separate from the one codified into the site's automatic calculations, courtesy of ElGuapo and others? Do I remember that? But I think I recall you saying it was very close?? Is the list above specifically for your own elo calculations, or is it for both sets? I have NO idea what a "CSV" format is, and I suspect 90+% of the folks here don't, either. So in general.......as I've noted before.......your lack of explanatory notes keeps me (and I'm sure others) from absorbing and utilizing some of your deep and detailed calculations. This is *not* a complaint - only again pointing out that your attempts to connect to the non-programmers among us is just going over our heads. Clearly your smorgasbord of ways to look at the data is wonderful, and several here can dig into them with ease. I can't, and I think that's likely true of most. But perhaps that's ok - your target audience is only the few who can digest your output, I think. I just asked what I thought was a simple(?) question for ElGuapo, suspecting the answer was "no" [See my post at Sat Aug 26 06:41:21 2023 above], but have yet to receive a straight answer - at least that I can understand.
And @ Punster - no need to go trod over that same ground again. We've been there and done that. I wasn't doing that. I was presenting a 'new' "theory"/suggestion - based on the supposed answer I would get to my question.
TN, i am astounded that you are not familiar with csv format
As I've noted before, my only programming experience was long ago - assembler language and Fortran. Got no clue what "csv format" is. No clue whatsoever.
TN, csv has nothing to do with programming. it's a data exchange format. so spreadsheets are also out of your realm as well?
Actually.....I used to be quite hot in spreadsheets - but that was also long ago. I was a whiz in Quattro Pro, and maybe Lotus, as I recall. Then I got good at Excel - but for various reasons, I didn't need to update that for a while, and my skills got 'stale'. So..........altho I *used to be good*, I'm not remotely current, and have never heard of "csv".
p.s. - For my purposes these days, Excel still works fine.
"First of all..............is it correct to say that you have formulated your own elo calculations, separate from the one codified into the site's automatic calculations,"
Right. But I like simplicity so I aproached it in a simpler manner (the basics part). But it still aproximates to the correct ratings. See "Games" tab for example. 8x4 level 5 is very similar to those in this site when sorted by ratings.
TN, so you download a .csv.... double click it.... and it opens Excel
i strongly endorse Hop's Hop Style Elo Info
its pretty much the way i would do it
its based on actual play data since 2021-09-13. its unfortunate that play data prior to that is not available
best feature is that if any tweaks to the elo calculations are required everything can be re calculated
TN — csv means comma separated values. Knowing that, you can get details from Wikipedia’s discussion: https://en.m.wikipedia.org/wiki/Comma-separated_values
==========
(Three posts were made while I slowly put together my response.)
csv files allow portability between applications. A table of data can converted to a csv file and uploaded in another application with rows and columns intact.
@ Hop - So help me understand the purpose of you developing your own calculations, if the result is so similar? I'm sure you had a good reason.
@ cellmate - interesting. MrFixit has many times told 'us' to get the files we want from the site, but he's never told us *how* - what is required. I'm assuming it's because he assumes that whoever *does* want them, can figure out how to get them - which is probably a reasonable assumption. I wouldn't have a clue how to begin - but that's ok; I have no desire to manipulate them for my own purposes. But it again highlights the gap between the programming-savvy here, and those of us who aren't. And again, I'm not complaining, just pointing out that disparity/gap. I'm again reminded of that study once upon a time that concluded about 85% of the burden of communication rests with the initiator - in general. People here have complained because I'm "too wordy", etc., which altho true at times, is because I represent the other extreme here: clarity through specificity - which often requires more verbiage. The use of acronyms unknown by many is another example of the communication gap herein, but that's (sort of) a separate issue.
My reason is curiosity. I wanted to see how the ratings would look like if all games start from an average rating. As you may know, in chess every player starts with an initial rating and over time the players reach their actual ratings.
The data is there for this kind of experimentations.
It also helps to see if we have good average ratings for the variants. For example the following is the averages for 8x4 level 5 to 12 (set in the code):
935,982,1023,1062,1093,1113,1248,1268, // 8x4
And the following is how they look like after processing the data at hand (they are very close except for level-5):
As cellmate pointed out above I may set 927.146 for 8x4 level 5 in the code and run it from scratch and see what happens.
cellmate says "it's based on actual play data since 2021-09-13". Is that not also true of the ElGuapo calculations? If not, how are they different? Some context would have explained that better.
And let me make sure I understand the different "starting point/initial conditions" of each method. ElGuapo assigned different ratings for the different variants, based on *perceived* difficulty?? (Not trying to put words on his keyboard; that's just my general perception from when this all started.) And then they've been adjusted, both by him, and by elo-adjusted play? Is that all right so far? *As opposed to* your method, in which ALL games started at 1500? And so then how are "good average ratings" evaluated? What is the standard? Is the standard ElGuapo's method, or your method?
And...............I don't want to make you all get bogged down in this; I'm just curious. There was FAR too much wrangling and discussion back at the (elo) beginning to absorb it all - at least for me. Way too many separate threads were ongoing at the same time. I don't want to make you waste toooo much time on this, because I'm not gonna get wrapped up in it - altho it may seem to some that I am already. :)
All that said.....................still waiting on an answer to my initial question..............
TN, are you asking if we maintain two separate ratings for games that appear in winnable play? We don't. You can see the game ratings on the game stats page. Each game has only one rating.
Not exactly. I know each game only has 1 rating. My question was: was "winnable-ness" taken into account when assigning the 'initial' elo rating? (Not counting the 4-5000 stuck on the unwinnables later.) I'm pretty sure not. Then my corollary would be: is that something that would be considered as a further refinement. Again, I suspect not. Just running it up the flagpole.......
I see. No, games were initially rated based on their win%. We're way past needing to refine the initial ratings though. Remember ratings have no memory, it's like a thermometer in water if that helps. It doesn't matter what the reading was before you put it in water.
Ratings have no memory , ok but it has an influence on the time needed to reach the right value
Your thermometer will take more time to reach 15°C if it starts from 99 rather than from 10
Consider, ElGuapo, this thought experiment......
Player A is playing 4x10, climate 5. According to the cumulative stats, 61+% of those are winnable (or shall I say *are* won - some of you all maintain the actual numbers somewhere not easily findable on this site). If said player is playing the regular version of 4x10, and draws game #160, for example, it is as of now 0/12, yet apparently is known to be winnable because its elo is listed as 1906.774. But said player would have no idea if the game he was facing was winnable or not, and clearly it's a difficult game. So he wouldn't know how much time to 'waste' trying to find a pathway, realizing that his quest might be futile. (As a certain significant percentage are indeed unwinnable.) There's a cost/benefit analysis that by necessity figures into how much time to spend on such games. There's a "diminishing return" concept at play.
Player B is playing 4x10, climate 5, but is playing the "Freecell Lite" version, also known as W4x10. He may also draw game #160, but will KNOW it's winnable if only (s)he takes the time (or takes the deck of cards) to crack the case. This knowledge represents an automatic advantage over the game. Yes, as joey, Punster, and others note, "it's still the same game". While true enough, the simple knowledge of having that advantage of "I'll win with enough effort and/or time (and/or cards)" makes a *huge* difference. It's a clear advantage.
So...........because of the way things were set up, player B will get a larger bang for his buck (reward/effort ratio) than player A, even tho they played the same game. Player A had a harder road to travel, having first needing to determine if the game was even possible to win.
So............to reiterate my point...................since the games are known to be either winnable or unwinnable (among the variants with both flavors available), it would have seemed 'reasonable' to include "winable-ness" into the initial ratings, in addition to just winning percentage. I know that would have been an extra step, and not knowing how the machinations of it all came together, that may have been a prohibitively long delay, as the process took quite a while as it was. I do fully realize this is kinda a minor point, in the grand scheme of things, and yes, general win/lose play will eventually work that out - in the very, very long run. But that's why I chose 4x10 for this example - the earth may no longer exist when the 4x10s get enough play to smooth that out.
But.............you've answered my initial question, that it was not taken into account. Thank you for your time.
Well, I knew it was way higher than 61%, but didn't know it was 99+%. And yes, I knew the vast majority of winnables were 'already' winnable. That mitigates my point a bit, but it still remains, imo.
Gee, TN, why do you SO MUCH enjoy being so obnoxious, ?
You may be just "running it up the flag pole, but I will never salute!
Tony, you are extraordinarily selective in declaring who is obnoxious.
I thought personal attacks were not allowed on this discussion board?
Really? TN's been attacking me for years, imo!