Kaya.gs wrote:...
I scanned over the Redux. I also used to believe Fischer was the best timing system for go. I literally hate canadian, i think it has a serious issue with players losing before the time is up (like having to play 20 moves in 20 seconds). And byo-yomi is very inefficient, because time management makes actually the game longer and more clumsy (reading yose before answering an atari is super distracting, just to be able to gain a few seconds).
The problem with Fischer is related to what you mention that 40-60% of the time is spent on a few set of moves ,classicaly L&D situations or turning points.
The problem in Fischer is that you have the lowest minimum time of all time systems. So if you are depending on your bonus time, and one of those hard situations comes up, you will have less time to deal with it than with ANY other time system.
For fast paced games i have no doubt Fischer is best, period. But reality will show us
.
I think I understand what you are saying about Fischer having the lowest minimum time. It is because people expect byo yomi or canadian (or Bronstein but we don't have that on our servers either
) to take less time due to poor usage. But if you plan a tournament that way you will run into those better at time management who use virtually every second of their allowance and overrun the planned round times.
Kaya.gs wrote:Now leaving the "competitivity" of hourglass and into the ratio:
I think its a very good feature of the time system that time management is less of a skill. Now the drawback you mention(that of course bears a lot of weight) is that it will tend to move at the speed of the fastest player.
Being more strict, hourglass is more a 1:1*n ratio with n depending on the Differential time (how i call its "main time").
Someone playing 2 seconds faster than their opponent in average, needs 240 moves to make a 4 minute difference.
(Calculation: A plays, B plays, and B gets 2 more seconds. That means he needs 120 double-plays to make 240 seconds).
So if someone takes average 10 seconds, and his opponent 12 seconds, its likely that a 5-10 minute hourglass would be more than sufficient for the vast majority of games.
What the slower player is subject to is vulnerability to a difficult situation he has to spend a lot of time to think of.
Remember that we added a minimum time to prevent corruption, so you always get at least 10 seconds. If your opponent thinks his average (12 secs) , you will get 22 seconds.
My conclusion so far is that Kaya.gs's Hourglass reduces the effect of the ability of time management, and provides pressure to the slower player mostly in critical situations.
Let's look at some examples. I could still find the games that I used to write Redux so all the timings were already in my old spreadsheets. This allowed me to quickly make some simulated hourglass timings.
All the following graphs show seconds on clocks on the vertical axis and game move numbers on the horizontal axis. In all cases I simply used the actual game timings that I had from the old files to simulate the play under hourglass timing. Whenever time was not available on the clock, the play was truncated to preserve 1 second remaining time. No attempts at active time management are assumed (e.g. use of additional time on the move following a truncation). In all the simulation graphs we see:
- B Clock/W Clock, representing an hourglass clock with no delay, and
- B Clck De/W Clck De, representing an hourglass clock with a 10-second delay.
Let's start with an easy one - a Canadian 25 stones in 3 minutes example:
Both players played so quickly that I had to use 120 seconds for the hourglass settings before the clocks even approached their limits. Neither player dominated this game. Note how the first 40 or so moves were all made in less than 10 seconds so no time moved on the B Clck De/W Clck De lines.
Next another good showing for hourglass:
In this 60-minute absolute game a ten-minute hourglass (300 seconds on each clock to begin) would have performed well. The players exchanged the position of holding the majority of time with Black ahead on the clock in the early going, White pulling clear around move 110, and then Black coming back for the end of the game. White hit a constraint on a single play around 180. In the game White thought for a little over 7 minutes, but this would have been truncated after about 3 minutes and 20 seconds in our simulation.
Something a little more problematic:
This is a second 60-minute absolute game with a 10-minute hourglass simulated. Why do I say it is problematic? Here are the actual clocks from the game file.
In the actual game the players used their time in very similar ways, as shown here. In total there was a cumulative difference of only 16 seconds between them at the end of the game. Yet in the hourglass simulation, White is constrained on several moves and loses 104 seconds versus the actual game. Personally, I am not so happy with this result considering how tight the pattern was on the actual game clocks.
Hmmm...
This is where it gets nasty. This is a third 60-minute absolute game. Obviously Black is dominated by White and ends up severly constrained in this game. Black losses 21.5 minutes in the simulation and over 19 even with the delay. So what, isn't that what is supposed to happen? Well, yes it is. However,...
Here are the actual game clocks. Note that White used much less than the agreed time. This is the situation that I believe is always on the table as the negative case with hourglass. That the actual play is different than our expectation.
One might say that we can fix this by increasing the time in the hourglass. That is correct, but probably not as easily as people would expect. In this case:
Code: Select all
(all times in seconds)
Hourglass Black lost time Black lost time with Delay
600 -1,291 -1,163
1200 -991 -886
1800 -691 -577
Even a 30-minute hourglass (15 minutes on each clock to start) does not really protect against a surprise.
.
