Blitz, Live, Correspondence thresholds (again)

:heart:

4 Likes

Auto correct hit.

I think that’s a reasonable calculation, but I think it’s misleading to group 30s byo-yomi only games with games that have both 30s byo-yomi and a comparatively large amount of main time which is the point.

The actual calculation must be somewhere, but lets assume it’s simple and similar to yours. Take the game time settings and assume the total number of moves is about 240 and the players use the max time available (just because I think people want to know what’s the longest I’ll potentially spend, rather than the quickest, which could involve early resignations, and players playing nearly instantaneously - much harder to estimate and depends on the opponent)

Normal ~ 20mins +5x30s. The game at worst could take up to 40mins (imagine 20 mins for the first two moves each) + 23830s+240s (using up the extra byo-yomis somewhere) which is a bit more than 2.5 hours. Per player the average move would be (2060+120+238*30)/240=35s/move. This must be close enough to the real calculation.

Blitz ~ 5mins +5x10s. The game would be more like 10mins+23810s +80s ~30 mins. The average time per move for each player is about (560+40+238*10)/240 ~11 seconds per move. Again this is probably close enough to the real calculation.

My guess would then be that when deciding if a game is Blitz or Live it takes the longest time the game could last divided by 2, or equivalently the longest time one side could spend on their moves, and then divides that time by some average number of moves for that board size, lets say 240, and that’s the average time per move. Then if it’s above some threshold like 12s,11s,10s it moves from blitz to live.

I’m not really sure what it would be from live to correspondence. Maybe if it’s taking longer than a day to complete, like something that would be possibly more than one sitting it would be correspondence.

Oh I take a break from writing to do something and come back and we have something far more definitive than my speculation :sweat_smile:

Although that might explain why the current Blitz settings are being counted as live, since I guess they’re averaging like 11s per move.

Edit: even looking back at some sitewide Blitz tournament games with 2mins +5x10s are being classified as Live in the game history, and if I calculate (260+23810+40)/240~10.5 seconds which maybe makes sense.

1 Like

It seems like blitz games are blitz and live games are live :crazy_face:

Oh so it’s supposed to be fixed for new games? And is it still strictly 10s or less per move on average to count for blitz?

Not sure. The takeaway I got is that it is not generally possible to determine if a game is blitz based on the time settings alone.

1 Like

Thanks Flovo! That’s for Fischer time, I take it. So with an increment of 5s the base time which is at the threshold between Blitz and Live should be 20 mins. (Which I guess, based on 240 being quite high, might be 10 mins main time for each player, not 20 mins.)

More 9x9 speculation: I saw for 9x9, 1.5m+5x9s byo-yomi, it registers as 10s/move on the graph, and ranks as Live rather than blitz. 90s+45s = 135s, though as @gennan pointed out, the 9s being max time per period, the average move will be a bit less. For both players: 180s + 90s = 270s, thus I would guess the number of moves expected for 9x9 is less than 30. :thinking:

Oh yes, when you do the math using the times stated in the time settings you have to divide base times by 120 moves in the case of 19x19 (and 30 for 9x9 I guess).

For Byo-yomi the period time takes the role of the Fischer increment, so in your example you have 90s/30+9s= 12s which is far more than the 10s lower bound for the live label.


Btw. the distinction between blitz and live is arbitrary and cosmetic. The difference between 9.9s/move and 10s/move is minimal, but resulting in different labels. These labels are only used to draw some icons in the game history and to calculate some unusable breakdown ranks, so there is not much to gain from knowing the categorization details.

Yea, I must admit my interest started by being “too engaged” with my ratings table. :slight_smile: And knowing where the threshold is doesn’t really matter: make my blitz games nice and fast, make my live games have some time to actually think (so nowhere near blitz), then I’ll have some clear record of how I’m doing on faster vs slower games! Skirting the threshold with live and blitz would be totally silly.

And for correspondence: I’ve found 4h per move, or even 8 hours per move, is too little - it takes over my life too much if I need to be checking that regularly. :slight_smile: So: I’ll stay away from the thresholds in the future. Just need my blitz games to be “blitz enough” so I can make some use of the separation in record keeping.

I hope this is the right thread to post this in (sorry if it isn’t!). I noticed that the games I find using the blitz automatch button are being counted as live games instead of blitz in my game history. My last two games were created with the blitz automatch button, one with Fischer and one with byo-yomi, and both show up in my history as live. I don’t know if it’s the automatch time settings or the blitz thresholds that need adjusting, but it seems inconsistent that games created with the blitz button aren’t counted as such.

1 Like

It’s a Known Fact that the distinction is not handled consistently.

That (and this thread) are only one of a string of comments/reports on this.

I think that @RubyMineshaft went as far as he could fixing it in the web client, but I think a proper solution has to come from @anoek in the back end, and this would surely be well down the priority list (due to other far more exciting and useful stuff in the pipeline).

2 Likes

I think if it’s a quick fix, and since it’s related to the most recent update, namely the ratings, I feel like it should be bumped up the priority list a bit.

I would go as far as saying it’s a bug in the new ratings update in that it also retroactively misclassified old games which were already counted as blitz and bunch them into the live ratings also, not just new games.

It’s not a quick fix.

1 Like

There is a lot grey between each of the subjective time labels. No matter where you draw the line, there will be bleed over of time settings between all of them.

I think this is part of why it’s not a quick fix. I think that a solid line - a clear definition - can be laid out, but only with a fair bit of thought, and even when that’s done you have to go through all the places where it’s band-aided over and make it all consistent. As I said, not a quick fix.

1 Like

I mean we can debate forever the borderline between live and blitz, but at the very least the blitz AST’s and the Blitz auto match button should actually be treated as blitz games in the ratings.

Maybe the quick fix is for the front end to relabel/change the display blitz ASTs and the blitz auto match button to say something like “fast live” or “rapid” instead.

2 Likes

I’d played a 1 x 10 sec 9x9 , which is treated as “live”. Not easy to play a blitz tho

1 Like

I actually wonder why that wouldn’t be treated as blitz.

Like in these types of calculations, doing it similarly wouldn’t it be 0/30+10s=10s which should be blitz, unless there’s a strict < symbol somewhere in the code instead of <=?

Yes, blitz is strictly less than 10s.

I think I stand by this then, since the site (automatch/AST) blitz settings aren’t blitz.

2 Likes