Yeah as far as I know, changing automatch to Fischer only is already decided, and it’s just a matter of choosing main/increment/cap for each speed.
anoek is notorioisly slow to pull the trigger on these kinda big changes. This length of waiting is nothing new here. It’s probably in the pipeline, he just likes to try his best to get things as close to right as possible with the first swing.
What?
I thought there was no consensus on this!
Only 33% of people support only fisher deleting byo-yomi, I think this is inaccurate!
Perhaps “decided” in the same way that a Japanese match with triple-ko is “decided” 
A poll is not the end of the discussion, but that reading is a bit misleading. 85% were in favor of one time setting deleting the other. Among those, Fischer was the most popular choice.
With 50% more votes than byo-yomi
50% sounds like a lot, but it’s actually just 60% vs 40%.
I don’t think such a gap is enough to remove byo-yomi and adopt fisher.
as said above, 85% voted that having just one time setting instead of two was preferable. That’s the reason one of them should be removed.
The statement of Fischer having 50% more votes than byo-yomi was only as a selector for that action, not the sole justification for it.
You yourself voted in favor of keeping just one time setting and deleting the other. I think it’s arguing in bad faith to come here now and oppose that just because the side you voted for is losibg. You were happy enough for Fischer to be deleted, so why not byo-yomi?
Sorry, my English may not be very good.
If I misunderstood, I apologize.
Is this an accusation against my character?
I thought we were all talking about how to make OGS better.
In fact, it’s none of my business how OGS’s live game is, I only play correspondence games.
Anyway, no matter what, I don’t plan to play live games on OGS at all, I only play on FOX.
I am just worried about setting the automatic matching live game to fisher and deleting byo-yomi, because this may make the escape phenomenon more serious.
I think changing the automatic pairing time setting is an important decision, and 60% vs 40% is not an overwhelming ratio. I was wondering if it was a bit hasty to make a decision like this, so I brought it up.
I’m just worried that wrong decisions will harm the OGS environment.
OGS is a good website, I hope it can be better and don’t want to see it get worse, so I put forward my ideas and discuss them with everyone.
If such enthusiasm is malicious, then I apologize and I shut up.
The bad faith accusation comes from the question of whether you’d apply this high threshold had Byo-yomi won out. I don’t think it was meant to be an attack on your character, although I can understand how you might perceive it that way.
Continuing out the debate prevents progress too. The right time to argue for a high threshold for change would be before the results are out.
Do you have evidence that Fischer games are more susceptible to escaping? And to what extent?
Good enough for Florida and Poland 
Great! I think we’re all trying to do that here.
Aside: linking the poll because it’s kind of buried.
If you just ask me this question. No, I won’t.
The reason is that I haven’t read any downsides to adopting byo-yomi from this thread, so I have no reason to advocate not using byo-yomi, if byo-yomi wins out.
But I think fisher is harmful because of the escapers.
This phenomenon is not obvious with Dans and SDK, because they usually run out of time and compress the time to the last second.
But you have to consider DDK and TPK players.
Usually DDK and TPK play so fast that they never run out of time.
If we use 5m+15s, assuming a game has 200 moves, each player will have 100 moves.
Then each player will have a total time of 1800s.
Do you think DDK and TPK will take so much time? I don’t think so.
Games on DDK and TPK can be over in 20 minutes, and are often shorter than that.
That is to say, on average, each player will have 1500s left.
Then, when the escape occurs, the victim has to wait 25 minutes!
This is definitely something that can happen, absolutely!
Yes, I have no evidence, or I have no data, and I don’t have the ability to make statistics.
This is all based on my imagination, but I’m not the only one in this thread to mention the disadvantages of fisher facing escapers.
However, I can guarantee that if I were a live player and something like this happened to me once (waiting 25 minutes to escape), I would never want to play again.
I know 60% is an absolute majority.
However, for some major decisions, an absolute majority is not enough.
For example, some companies require the approval of more than 67% of shareholders for their major decisions.
And changing the automatic pairing time setting, I think this is a big decision.
Maybe 67%? 75%? 80%?
Would it be safer if the number of votes was several times the number of objections?
Again, I bring this up because I see potentially dangerous decisions on the horizon.
If the decision adopted has no hidden dangers, why would I object?
You can’t say that I have bad faith.
If you are playing on IGS and your opponent forgets to take the dead stone before confirming the score, are you obligated to help him take it without confirming the score directly? You can choose to do this, but you are not obligated to do so.
That’s a fair point and in my view would, at least, be an argument in favor of reasonable caps. Such as capping at the main time (so, here, 5min).
I think you’re reading too much into this though.
Only if the eacaper gets up and walks away from their computer with the game tab still open. This is not typical. Most escapers navigate away from the game to start a new one. In which case the maximum wait would be 5 mins as it is now, regardless of clock options.
+1 to main time = cap
Well, I really didn’t expect to have a time cap.
Although the time cap feels bad (because it seems to encourage you to use up time instead of saving it. However, byo-yomi is more serious), it does solve the problem of needing to wait 20 minutes for escape to occur.
I know it’s not typical, but the devastation it can cause once is huge.
And if it is a byo-yomi of 5min+3x30s, usually when the escape occurs, the victim only needs to wait three byo-yomi, which is 90 seconds, instead of waiting for 5 minutes.
Re Fischer / Byo-Yomi, it seems very obvious to me that we should try Fischer as the default, especially since we have all the plumbing already in place to have both as options and allow people to prefer or require Byo-Yomi if they so choose, which I think we should keep in place for the foreseeable future since I recognize a significant part of the population will likely still prefer Byo-Yomi for some time to come.
60/40% to me is a very strong signal because of the fact that Byo-Yomi is the de-facto standard everywhere, yet we have a majority here saying that we should make a switch. Folks are generally resistant to change, so the fact that we had a notable majority overcome that change inertia and vote for Fischer, combined with internal pressures at the EGF/AGA and various tournaments to move to Fischer as well, indicates to me that collectively the community is moving in that direction and the site should do so as well.
I’ve spent a good deal of time crunching through the past several years worth of games to try and tease out what the best set of Fischer automatch time settings might be, and to double check that the proposed Byo-Yomi times above are optimal, which I was not surprised that they are.
Below are the highlights of my findings and my proposed settings for 19x19 games (9x9 and 13x13 to come after we finish working through 19x19). Feedback would be highly appreciated.
Analysis notes:
| Speed | Settings | Game duration | Moves | Avg. Move time | Samples | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Avg | 50th | 90th | 99th | Avg | 50th | 90th | 99th | Avg | 50th | 90th | 99th | |||
| Blitz | 30s + 5x10s | 11:02 | 10:36 | 16:40 | 21:58 | 193 | 189 | 278 | 325 | 3.44s | 3.36s | 4.44s | 5.43s | 5243 |
| Rapid | 5m + 5x30s | 21:41 | 19:59 | 34:23 | 49:42 | 208 | 210 | 286 | 329 | 6.31s | 5.85s | 9.37s | 13.05s | 39042 |
| Full | 20m + 5x30s | 32:11 | 28:21 | 49:09 | 67:07 | 210 | 216 | 288 | 328 | 9.39s | 8.13s | 13.46s | 19.49s | 162045 |
Blitz: As noted the the original thread creator @AfricanRhino17, 5m+5x10s feels a bit too slow for a blitz game since there’s so much main time. With an expected game duration of 15:33 - 23:10 and an average move time of 5.36s, it resided somewhere between Blitz and Rapid. This probably wasn’t the worst choice when there was only “Blitz” and “Normal”, but as we introduce a new Rapid category, having a faster Blitz makes sense.
The choice for 30s+5x10s specifically from the observation that most popular byo-yomi blitz setting is 0s+5x10s with a close second coming in at 1m+5x10s. 30s+5x10s splits the difference, and has a certain alignment with one of the favorite Fischer blitz settings of 30s+3s, so it’s a compromise between the two most popular settings and pairs well with our proposed Fischer settings.
As an interesting side note, despite most of the game being played in overtime, the average game duration for a 0 + 5x10s game is 10:05 where as 1m + 5x10s is 11:19, nearly a minute difference just like the main time.
The choice of 5x10s vs 3x10s is simply based on popularity, the former being about 2.5-3x more popular.
Rapid: As noted by the thread, 5m+5x30s is a very popular choice, followed by 5m+3x30s. In terms of expected game times it fits nicely between the longer traditional 20+5x30s games and the new 30s+5x10s blitz games. There wasn’t really much question as to alternatives to this time setting, it is very popular with no real rivals.
Full / Traditional / Normal / Classical: Whatever we want to call this, these are the current default settings for a lot of games and seeing how it fits nicely in terms of a close to 10m increment in expected game time, I see no reason to change it. I did look at some other variations, such as 30m+5x30s (36:37 - 69:46) and 20m+5x60s (36:51 - 66:51), but neither of those settings were particularly popular in the custom game settings and I think 20m+5x30s fits better along side our our blitz and rapid picks.
The goal with picking appropriate Fischer settings to achieve similar cadence on average to their Byo-Yomi counterparts, but pull in the 90th and 99th percentiles to provide a more consistent and predictable game experience, particularly for blitz and rapid games.
| Speed | Settings | Game duration | Moves | Avg. Move time | Samples | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Avg | 50th | 90th | 99th | Avg | 50th | 90th | 99th | Avg | 50th | 90th | 99th | |||
| Blitz | 30s+3s | 11:02 | 9:56 | 14:14 | 16:40 | 197 | 196 | 279 | 319 | 3.65s | 3.10s | 3.36s | 3.61s | 2710 |
| Rapid | 5m+5s | 21:33 | 21:27 | 31:04 | 34:34 | 195 | 189 | 277 | 312 | 6.79s | 6.95 | 8.56 | 10.00 | 179 |
| Full | 10m+10s | 33:06 | 26:18 | 52:11 | 68:36 | 204 | 205 | 288 | 334 | 9.53s | 7.79 | 13.66 | 17.42 | 860 |
Blitz:
30+3s is a popular Fischer choice with a measured and theoretical cadence that matches well with our measured 30s+5x10s Byo-Yomi blitz time setting. A lot of options were considered, but some particularly noteworthy alternatives were:
| Setting | Notes |
|---|---|
| 3m+2s | A popular chess setting, but not popular for custom games (only 24 samples). The average length of those games were 11:14, so slightly longer than 30+3s, yet the 50th percentile for moves made was 157 indicating a higher propensity for timing out. Overall this seemed worse than 30+3s, but we don’t have a lot of data. |
| 30s+4s | Also reasonably popular for custom Fischer blitz games and had an average game time of 11:36, so could be a reasonable alternative, but 30s+3s is more popular and I think fits the desired cadence better. |
| 1m+3s, 2m+3s | These seemed reasonable, but were less popular than 30+3s |
| 30s+10s | A current default setting, this measured with an average game time of 16:35 but had larger upper bounds of how long games last. 16:35 also falls in-between what we want for Blitz and Rapid. |
| 5m+5s | This turned out to be way too slow, but perfect for Rapid. |
| 30s+5s | With an average game time of 14:15 this seemed to fit in-between our desired Blitz and Rapid target times. |
| 1m+5s | With an average game time of 15:22 this seemed to fit in-between our desired Blitz and Rapid target times. |
| 1m+6s | With an average game time of 15:02 this seemed to fit in-between our desired Blitz and Rapid target times. |
Rapid:
5m+5s is not a common setting. Much more common were 5m+10s, 2m+10s, 30s+30s, and 2m+15s, however for each of these while the average game durations were about what we wanted for a Rapid game, the variance in durations was a lot higher as they all have the significant possibility of starting to feel like a “fulll” as the time increment is greater than the average move time in a “full” game. 5m+5s provides a more predictable game duration, which I think is an important consideration for both Rapid and Blitz.
Full:
10m+10s is a reasonably popular time setting and aligns pretty well with 20m+5x30s. We diverge from blitz and rapid here in that our time increment actually exceeds our average move time, but I think in this case it’s appropriate to facilitate slightly slower than anticipated play while still keeping a reasonable cap on the games. Some other common time settings were 3m+30s, 14m+30s, 5m+60s, and 3m+20, however while the average game duration for those games was reasonable, the 90th and 99th percentile games notably exceed that of 20m+5x30s, which makes sense since there’s so much opportunity to bank an excessive amount of time with those time increments.
If the backen classification isn’t changed this games will show up as live games in the game history and ranking breakdown
Yeah we’ll have to fix that up. @dexonsmith has some pretty good plans in the works for a lot of ratings things and we’ll roll the adjustments and fixes in with that as we roll those updates out.
byoyomi
Fischer
I get the rationale of each of those rows from data collecting, but I can’t help but notice some overall imbalance in the overall end result.
Like the Fisher overtime increases by roughly a factor of 2 between categories, but byoyomi overtime doesn’t change from rapid to full. Wouldn’t byoyomi rapid overtime of 5x20s fit more nicely in between blitz 5x10s and full 5x30s, like Fischer rapid overtime is in between blitz and full?
And basic time increases by a factor of 4 from rapid to full with byoyomi, while with Fisher it only increases by a factor of 2. Wouldn’t 20m Fisher full basic time make a more balanced pattern? (or 10m byoyomi full basic time if that is too slow, or perhaps both 15m basic time, increasing by a factor of 3 from rapid basic time).
Here’s the data for those:
| Settings | Game duration | Moves | Avg. Move time | Samples | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Avg | 50th | 90th | 99th | Avg | 50th | 90th | 99th | Avg | 50th | 90th | 99th | ||
| 5m+5x20s | 17:40 | 16:47 | 27:07 | 38:25 | 208 | 211 | 285 | 326 | 5.16s | 4.89 | 7.36 | 10.26 | 6466 |
| 10m+5x30s | 24:40 | 22:50 | 39:04 | 54:39 | 208 | 211 | 287 | 330 | 7.25s | 6.64 | 10.73 | 14.94 | 67814 |
| 20m+10s | 38:32 | 36:32 | 63:37 | 84:49 | 216 | 220 | 294 | 360 | 10.91s | 10.55 | 17.08 | 26.08 | 174 |
I think 5m+5x20s and 10m+5x30s could be good choices too. They’d provide overall a faster play, but popularity wise it seems like folks really like 5+5x30s, and it’s somewhat of a standard it’d seem, being popular on other servers as well as our own custom matches. Here, 5+5x30s had 39042 custom games compared to 5+5x20s with “only” 6466, so about 6x more popular. So if we say that 5+5x30s is our desired Rapid setting, I think that provides some reason to go with 20m+5x30s over 10m+5x30s as the next step up, otherwise they’d feel a bit too close in terms of how fast folks tend to play in those games.