question about bbo hand generator
#1
Posted 2008-April-14, 20:16
A couple of days ago, I was sitting south and West was declarer at game for 24+ games in a row.
Several players in the East hand would actually leave the table, tired of being dummy after 2 or 3 hands. the games were bid on solid cards.
I am sure my partner did not enjoy having to be the opening leader twenty-odd times in a row.
I myself did not have much to do, as the hands I was dealt were not even suitable for defensive bidding purposes.
Also in tournaments, It is not uncommon to hear players inquire whether the game is goulash.
What are the odds of getting a single pattern of hands for long streaks at a time? Even for days?
#2
Posted 2008-April-15, 01:51
#3
Posted 2008-April-15, 02:29
#4
Posted 2008-April-15, 02:45
#5
Posted 2008-April-15, 03:55
brianshark, on Apr 15 2008, 09:29 AM, said:
No way, it's at best PSEUDO-random, which isn't the same...
#6
Posted 2008-April-15, 05:28
brianshark, on Apr 15 2008, 09:29 AM, said:
Yes. This issue comes up again and again. People keep seeing patters where there are none.
The random number generators used have been validated on millions of samples. The impression of a player who has played a few thousand hands has no weight.
BTW I would say that randomness and pseudo-randomness are the same but I suppose that's a philosophical discussion. For practical purposes it suffices that the distribution generated by computers are indistinguishable from the ones predicted by probability theory. Which may not be true for hand-dealt boards.
#7
Posted 2008-April-15, 05:58
Free, on Apr 15 2008, 09:55 AM, said:
brianshark, on Apr 15 2008, 09:29 AM, said:
No way, it's at best PSEUDO-random, which isn't the same...
Yeah, I was waiting for that retort.
I'm aware of the concept of pseudo-randomness and the inherent impossibility of true randomness. But I'm pretty sure BBO hand-generation is pseudo-random-enough for you to still be imagining any bias or pattern.
#8
Posted 2008-April-15, 06:16
helene_t, on Apr 15 2008, 11:28 AM, said:
brianshark, on Apr 15 2008, 09:29 AM, said:
Yes. This issue comes up again and again. People keep seeing patters where there are none.
The random number generators used have been validated on millions of samples. The subjective impression from a player who has played a few thousand hands has no weight.
Helene is correct.
When I first wrote the hand-generation aspect of the software (7+ years ago) I did some extensive statistical testing to make sure I got this right. This part of the software has not changed since then.
Normally when I receive reports like this I dimiss them as a matter of perception or as the sort of streak that should be expected to (randomly) happen. That being said, a streak of 24+ deals where West has the best hand at the table is so unlikely from a statistical point of view that Babalu has managed to get my attention (though I have to admit that it is hard for me to imagine any programming error that would result in behavior like what Babalu reports).
To Babalu: I have glanced at a record of all of the hands you have played during the past month and I have not noticed any particular streaks of "unusual" length. It is entirely possible that I missed something strange. If you can send me either BBO log file as an e-mail attachment (fred@bridgebase.com) or tell me what dates/times to examine in www.bridgebase.com/myhands I will take another look.
Fred Gitelman
Bridge Base Inc.
www.bridgebase.com
#9
Posted 2008-April-15, 07:23
But you will note that starting with taht deal E-W was declarer for 10 of them. the only one which was declared by south, deal 30318772, was actually an ill-advised balance of 2 spades, which I actually made by the skin of teeth.
So perhaps an 11 hand streak is not considered noteworthy statistically, but I will be tracking these carefully from now on, even if it is just a collection to demonstrate that perhaps Karapet Djoulikian is my distant relative.
If 12-14 streaks of hands are consistent with the laws of probability then it explains why it is possible that a 12 board tournament can potentially be stacked on one side of the table. If so, is there a tournament movement which could minimize such?
Thanks
#10
Posted 2008-April-15, 07:45
babalu1997, on Apr 15 2008, 02:23 PM, said:
No. It does not depend on the movement. If you play 11 boards, the probability that one side gets the majority of the strength 10 times is approximately 1% (if we neglect the possibility that the strength could be distributed evenly, i.e. 20 HCP vs 20 HCP). So it is bound to happen once in a while. And if you play a longer series of boards, the probability of a strike of 11 such boards is substantially higher. And of course, some pairs get less than their fair share of declarership because they bid less aggressively than others.
In principle one could bias the deals so that each pair gets the majority of the strength on at least, say, 3 of the 11 boards. But this would have a funny effect: if you play an 11-board tourney and noticed that your side had the minority of the HCPs on 8 of the first 10 boards, you know that your side has the majority of the points on the last board, a fact you could exploit.
So there is nothing to do about it. Try to appreciate the play with bad cards, or comfort yourself with the thought that even if you got bad cards on this deal, in the long run you will average 10 HCPs just like everyone else.
#11
Posted 2008-April-15, 09:33
NS EW
1♠ 4 3
1♥ 3 5
1N 24 25
2♠ 33 31
2♣ 6 5
2♥ 23 41
2♦ 13 7
2N 13 18
3♠ 23 23
3♣ 23 14
3♥ 28 25
3♦ 25 18
3N 88 104
4♠ 74 62
4♣ 10 5
4♥ 61 59
4♦ 6 6
4N 2 0
5♠ 8 9
5♣ 8 13
5♥ 3 8
5♦ 15 17
5N 1 0
6♠ 5 7
6♣ 1 3
6♥ 7 7
6♦ 3 4
6N 4 6
7♥ 1 0
7♦ 1 0
7N 0
Lets look at the results where you were south only. Here there was a slant of games for east west. They bid game or more 139 times while you bid it only 95. When you were north, NS bid 40 games+, EW bid 34. So while your were NS, there was a slant of bid games for EW (135 to 173). There was a smaller slant of games but towards NS when you were sitting EW (79 to 71). This could be luck of the draw, or as helene says, maybe your side bids game less aggressively than some of your opponents.
#12
Posted 2008-April-27, 05:35
Let me illustrate the annoying part of the randomizer in this area. The HCP totals of my side were as following: 21, 28, 19, 23, 26, 28, 23. This gives my side a nice average of 24HCP over 7 deals. In this case the annoying part was for my opponent, who obviously left the table... I know 7 deals isn't much, but almost EVERY time I play moneybridge I get similar situations!
When I looked further, I noticed that the direction in which I was sitting switches. The first 2 deals I was West, while the rest of the deals I was sitting South. If we calculate the average hcp for NS instead of my side, we will count an average of 21.43 which is statistically a lot better!
So although the randomizer seems to work ok, in moneybridge it's not working properly as it should. By not switching seats I think you can already get a lot of improvement.
#13
Posted 2008-April-27, 19:26
If your opponents get 3 solid games in a row in a Swiss Match and bid well, you are catapulted to the bottom along with all those who had good opps.
Sometimes there is no recovery.
Also, i would not know how to evaluate the card distribuyions, but i often say that trumps only break 3-2 in the newspaper. That is tru for the opps too.
The side effect of this is that it forces us to learn more of those declarrer play magic tricks and also false carding.
#14
Posted 2008-April-28, 02:25
babalu1997, on Apr 27 2008, 08:26 PM, said:
Some not so strange truths:
The better your play technique, the better your good luck becomes and the less bad luck you have.
The better you become as a defender, the more bad luck your opponents will have.
Luck is always there, but you must give it a chance to work for you.