[PrecisionL]

JLOGIC, on 2012-August-04, 08:42, said:

I do not know why you are so defensive. You can argue with emotion and appeals to non-relevant authorities all you want. You can even attempt to discredit me by pointing out directly or indirectly that I am not as good as fantunes, and thus rate to lose to them. That would be relevant if I was like LOL FANTUNES SUCK.

However, your op was very clear:



You even wrote CONCLUSION so I assume it was clear what I was talking about with my "criticism" I offered, that the conclusion was not based on math or logic. It is basically one big logical fallacy.

You see people attempt to use numbers in this way, but there is often a problem with causation. This is very common in many books, studies, etc where people try to analyze data.

I realize that the conclusion might have been your conclusion from data that bill jacobs offered in his book, but I took it to mean that it was a conclusion bill jacobs drew in his book. Whoever drew that conclusion, specifically that 42 % of bidding accounted for their gains is obviously wrong. Here is an example:

If my style is to bid scientifically, carefully catering to all possible slams etc, then I will sometimes find a good slam that the other table missed. Ok, great, I won the board with my bidding system/judgement, and that is factored in. However, how about the times that I do the same thing, and I give them lots of information to make the killing lead, or the winning defense. This is the tradeoff you make for bidding carefully rather than blasting when slam is unlikely. So now I got to the same game as the other table, but I went down and they made it. By the conclusion above, this would mean that my cardplay was inferior, but really it was my bidding that caused me to lose that swing, despite ending up in the same contract.

So the fallacy here is that if we get to the same contract, our system/judgement in the auction was irrelevant, and imps won or lost are based solely on the cardplay. See, that wasn't so hard! There are more things like that where the conclusion drawn does not logically follow from the data given. Ergo, the premise that because they win .67 imps/bd on hands where they open, and .38 imps/bd when they play the same contract, they win 57 % of their imps from superior cardplay and 42 % from bidding is just wrong. It is based on underlying bad logic/math, which is what I said.

For what it's worth, it's possible that fantunes are winning more than 42 % of their imps from bidding. It is possible they win less from bidding. I have no idea from the data provided, and I would not attempt to draw that conclusion from that data. I did not draw any conclusions from it, or offer whether the book was good or not, or whether fantunes were good or not, or whether I think their system is good or not, and what % of imps they win from bidding. I do not know those things. I do know that the conclusion given in the OP was ridiculous, which was the only thing I commented on.

The rest of your last post was pretty amazing and also filled with bad logic, but I'll just end it here.

Wow, I don't know what is going on here, I just posted interesting summaries of two recent books that do more than provide anecdotal evidence to support their conclusions and you suggest I am defensive and emotional.

Justin, you are taking offense where none is intended. I was just pointing out that you dissed my posting and assumed what my intent was without providing any constructive criticism or data to discount Bill Jacobs book's conclusion. I did not present my own opinions and therefore there was no logic to criticize.

[bluecalm]

Quote

However, how about the times that I do the same thing, and I give them lots of information to make the killing lead, or the winning defense.

What I tried to do is to take all hands played by given pair and calculate double dummy result after the 1st lead and then compare it to minimax on given hand. This way you account for revealing bidding and it *SHOULD* more or less have similar EV as real level of bidding. Variance is obviously huge though and not all things are factored (ie first lead is, subsequent defense isn't).
I realize this method is far from perfect but at least it's a try as opposed to:

Quote

Ergo, the premise that because they win .67 imps/bd on hands where they open, and .38 imps/bd when they play the same contract, they win 57 % of their imps from superior cardplay and 42 % from bidding is just wrong. It is based on underlying bad logic/math, which is what I said.

Which I agree is just nonsense.

Also Fantunes played much more hands than 2700 on vugraph. I have 7119 hands right now on my computer and I guess that's not all of them as I haven't downloaded 2012 (and end of 2011) hands at all.

Quote

And what about the book: WHY THEY WIN by Stephen Cashmore & Justin Corfield, Scotland, 2008. Their conclusion based on three 24 board IMP matches is that Bidding Judgment was responsible for the majority of the IMPs exchanged. One team involved Zia & Robson.

Laughable sample size and probably bad methodology. Every argument like that I saw for bidding judgement being the most important thing at high level suffered from the author not understanding probability and being very result oriented ("uh-oh they bid 3NT here which is 50% contract but 4S is only 15% contract and 3NT maked so +12 for judgement!")

[PrecisionL]

bluecalm, on 2012-August-04, 13:21, said:

Also Fantunes played much more hands than 2700 on vugraph. I have 7119 hands right now on my computer and I guess that's not all of them as I haven't downloaded 2012 (and end of 2011) hands at all.

Yes, Fantunes played more than 2700 hands on BBO, what I posted was 2723 hands that they opened the bidding on.

WHY THEY WIN as I understand the book did the analysis you propose - yes it is a small sample, and probably is not statistically significant, but both books present knowledge that is food for thought.

So, who has better information / data to present that would cause those of us who are mathematically / statistically minded to consider other explanations?

[bluecalm]

Quote

but both books present knowledge that is food for thought.

Imo the author's argument is weak total junk. I don't want to quote the whole chapter here but his reasoning isn't even a serious try to quantify anything. He makes an argument that on hands with the same contract "cardplay swing" is 1.76 on average while average swing on any hand is 4.1 so from that follows that cardplay accounts for less than 50% of swings.
What is worse he then goes from that to conclusion about how much bidding contribute to total edge of given pair.

I mean, can I have some of what the author is smoking ?

One fact which almost all authors ignore is this:
Just because something contribute to most of the swings doesn't mean it's significant part of pair strength. It might be just variance. Of course the biggest swings come from bidding but the difference in EV is often very low. If you bid 52% slam or not or if you bid thin game which happen to make opposite actual partner's hand (but wouldn't opposite many other possible hands) doesn't matter much.
Reasoning as: "average cardplay swing is 1.76 and average swing is 4.1 so bidding contribute to more than 50% of pair strength" isn't "food for thought" it's just nonsense. Misleading nonsense to be sure.

[Free]

PrecisionL, on 2012-July-25, 19:20, said:

CONCLUSION: Superior card play [both by declarer and the defenders] accounted for just over 57 % of the IMPs won, while superior bidding (bidding judgment & system) accounted for 42 % of the IMP gains.

From this, you might as well conclude that their opps' mistakes account for only 1% of the imps won. WOW!

[AdamL3]

laststandb, on 2012-August-01, 11:35, said:

Fantunes is not GCC Compliant because it uses Transfer Responses to it's opening 1C bid, and also it fits 4441 hands into the 1NT bid.

I'd imagine a GCC Compliant Fantunes system would move 4441 Bids to 1D and 1C, and either change the 1C bid to 15+, or simply remove the transfer responses.

I'm interested in the system and I'd be interested if we (the forum) had an effort to try to convert Fantunes to GCC

For a while now I've been playing a version of Fantunes that's modified for ACBL GCC, as follows:

- 1C must be 15+ (and 2C therefore is 11-14)
- Since 1N cannot include 4441, we open 1D unless the singleton is in diamonds. With the 4=4=1=4 we have to open 1H. (After a 2-level response, we play that 2N by opener shows the minimum 4441 with singleton in partner's suit) and is NF.
- 1M-2C non-GF is not permitted on less than three clubs, so we have to bid 2D on balanced or major suit raise hands with fewer than 3 clubs and at least 3 diamonds (and we use 1M-3N to cover GF raises without three in either minor). 1M-2D isn't a very comfortable auction as a result, but we manage.

Also their 1D-2D on a weak reverse-flannery shape isn't allowed, but that isn't really that important to the system as a whole.

[helene_t]

I think all this criticism of the stats is a bit over the top.

Sure, the figure 57% is not likely to be particularly accurate. It is biased by a lot of issues. For example, when Fantunes do not open the bidding in first seat, the negative inference available to both partner and opps is different from the inference available when a different pair fails to open in first seat. Also, there is a correlation between opening the bidding and declaring, and it could be that their declarer play is better than their defense. Etc etc etc.

But my feelings are that those biases are probably not so strong as to make the data completely useless. I still find it intriguing to look at such numbers even if is not not entirely clear what to make from them.

[bjacobs]

Hi,

The GCC restriction on transfer responses to 1C is a big problem for Fantunes. Without them, the whole system is thrown out of kilter. Given the growing popularity of transfers after 1C, one would hope this restriction will not last forever.

I had some experience with this in the Akuba Cup, the Pairs event that is part of the NEC Cup. We were told not to use the transfer responses, and as it happened it cost us the event. We had one board where the auction started 1C-1H (natural, perforce), and the resulting 4H contract was played from the wrong hand. This cost many matchpoints, and we came second by about an eighth of a top.

The prohibition on the Fantunes 1major - 2C response is probably also a killer. The restrictions on the 4441 shape would be less of a concern (you improvise).

Cheers ... Bill Jacobs

[bluecalm]

Quote

But my feelings are that those biases are probably not so strong as to make the data completely useless. I still find it intriguing to look at such numbers even if is not not entirely clear what to make from them.

But those numbers have nothing to do at all with how much the system is worth.
Look at this situation:
Pair A never bids 50% slams and always bids 50% games. Their cardplay is perfect.
Pair B always bids 50% slams and never bids 50% games. Their cardplay is such that they lose makeable contract one time in 10 and it costs 10 imps every time.

Now if we play a lot of hands with 50% games and slams vast majority of swings will come from bidding not cardplay. In fact cardplay is as good 9 times out of 10. If you apply author's argument here you will end up with some non-sense like bidding contributing significant % to total pair A's edge. Still, cardplay is 100% of the edge of pair A.

The author just measures the wrong thing. When you try to approximate something at least give argument which makes some sense and works on simple example data.

[semeai]

bjacobs, on 2012-August-07, 18:40, said:

Hi,

The GCC restriction on transfer responses to 1C is a big problem for Fantunes. Without them, the whole system is thrown out of kilter. Given the growing popularity of transfers after 1C, one would hope this restriction will not last forever.

I had some experience with this in the Akuba Cup, the Pairs event that is part of the NEC Cup. We were told not to use the transfer responses, and as it happened it cost us the event. We had one board where the auction started 1C-1H (natural, perforce), and the resulting 4H contract was played from the wrong hand. This cost many matchpoints, and we came second by about an eighth of a top.

The prohibition on the Fantunes 1major - 2C response is probably also a killer. The restrictions on the 4441 shape would be less of a concern (you improvise).

Cheers ... Bill Jacobs

Welcome to the forums! Thanks for writing the book, and for your thoughts & anecdote here.

bluecalm, on 2012-August-07, 19:50, said:

But those numbers have nothing to do at all with how much the system is worth.
Look at this situation:
Pair A never bids 50% slams and always bids 50% games. Their cardplay is perfect.
Pair B always bids 50% slams and never bids 50% games. Their cardplay is such that they lose makeable contract one time in 10 and it costs 10 imps every time.

Now if we play a lot of hands with 50% games and slams vast majority of swings will come from bidding not cardplay. In fact cardplay is as good 9 times out of 10. If you apply author's argument here you will end up with some non-sense like bidding contributing significant % to total pair A's edge. Still, cardplay is 100% of the edge of pair A.

The author just measures the wrong thing. When you try to approximate something at least give argument which makes some sense and works on simple example data.

It seems as though you've missed that the book and the OP are discussing net imps/board, not number of swings or total imps exchanged. The bidding for pair A and pair B have zero net effect on imps/board in the long run.

I have to agree with helene_t that the data are somewhat interesting, if not particularly pure. Your (bluecalm) great thread from last year comparing declarer play, defense, and bidding + lead for various players to double dummy has statistics closer to the sort one would like to be looking at for the contribution of bidding vs card play, but these aren't nothing.

I'll agree with various other posters that the data likely don't have much to do with the superiority of the system, but rather that the best one should hope for from this data is some sense of the value of the whole of their bidding (system & judgement). Justin's good point that bidding and cardplay can't be separated this simply (and other similar concerns) does prevent easy access to this "value of the whole of their bidding." However, if we can make some guess about how revealing their auctions are compared to their peers (and other things), maybe we can draw some conclusion, if only an (approximate) inequality of the form "the value of their bidding as a whole is likely [greater than / roughly equal to / less than] such-and-such imps/board."

[ulven]

I’ve compiled vugraph data, starting 2006, up until (but not including) the Dublin Europeans for Fantunes (5097 deals) and used filtering in Double Dummy Solver (http://www.bridgecap...downloadDD.html)

Some stats that might be interesting:
They are then, compared to the other table in play (all deals are team play): +0.64 imps/board.
Fantunes open (2423 deals): +0.71
Opps open (2652 deals): +0.58
Fantunes declare: +0.52
Fantunes defend: +0.77

Fantoni open (1185 deals): +0.27
Nunes open (1238 deals): +1.13
They open, auction uncontested: +0.42
They open, opps overcall: +1.03
Fantoni overcalls (737 deals): +0.20
Nunes overcalls (714 deals): +0.69

Looking at 1NT-opening:
641 times: +0.66
When LHO overcall 1NT (149 deals): +0.42
RHO overcall (132 deals): +1.30

Can go on here for a while…

[bjacobs]

semeai, on 2012-August-07, 20:38, said:

Welcome to the forums! Thanks for writing the book, and for your thoughts & anecdote here.

It seems as though you've missed that the book and the OP are discussing net imps/board, not number of swings or total imps exchanged. The bidding for pair A and pair B have zero net effect on imps/board in the long run.

No, I hadn't missed the discussion, it's just that I can't think of much to contribute.

I did some calculations that are accurate if BBO Archives and Excel are to be believed, and drew a couple of conclusions from the results. Nothing in this thread makes me want to back off from those conclusions, although I will admit that there are some points made here that I simply don't understand.

The analysis that I did is fairly important to me, because there are probably people in the world who are thinking: "Fantoni and Nunes are such great bridge players: think of what they could achieve if they played a normal system!" I am totally convinced that their system is one of the reasons for their greatness, rather than something that holds them back. Most of that belief stems from my own experiences with the system, for which I kept substantial statistical records that I chose not to publish in the book, although they can be found elsewhere on the internet if you are a REALLY good searcher

Cheers ... Bill

[glen]

bjacobs, on 2012-August-08, 00:51, said:

... I kept substantial statistical records that I chose not to publish in the book, although they can be found elsewhere on the internet if you are a REALLY good searcher

I'm a REALLY good link clicker

ulven, on 2012-August-08, 00:50, said:

Fantoni open (1185 deals): +0.27
Nunes open (1238 deals): +1.13

Nunes Revealed!

ulven, on 2012-August-08, 00:50, said:

Can go on here for a while…

Please continue and post stats by opening.

[han]

Bill, it is wonderful that you have entered the discussion here. I haven't read your book yet, but I am very happy that you wrote a book on this topic and I'm looking forward to reading it.

PrecisionL quoted this from your book:


<sub>2723 deals, net IMPs won = 1817, or 0.67 IMPs per deal on deals Fantunes opened the bidding.
1676 deals, net IMPs won = 645, or 0.38 IMPs per deal where the contract was the same at both tables.

CONCLUSION: Superior card play [both by declarer and the defenders] accounted for just over 57 % of the IMPs won, while superior bidding (bidding judgment & system) accounted for 42 % of the IMP gains.</sub>


This conclusion is too strong for several reasons. In my mind the biggest problem with the conclusion is that IMP scores are relative to the other table. Fantunes have often played with extremely good teammates, who are among the best defenders in the world. Therefore the IMPs gained by Fantunes is partly because they received (on average) easier defense than the pair they are compared with.

You might say, but this is true for all the hands, and there is still the difference between the hands where the contract is the same, and the hands where the contract is different. This is true, but good defense matters most on the hands where the contract is close, and not on boring flat hands. And boring flat hands are exactly the kind of hands where both tables are likely in the same contract. So, you would expect that good defense matters more on the set of hands where the contract is different. How big is this factor? I have no idea, probably nobody knows, which makes it hard to draw conclusions as above.

But as Helene says, the data is very interesting, and it seems likely that Fantunes' bidding (system + judgement) is extremely good.


_{I am totally convinced that their system is one of the reasons for their greatness, rather than something that holds them back.}


You play the system yourself in a high level partnership, and you have often seem them play. Your opinion on the system should therefore be taken seriously. Unfortunately, you did yourself a disservice by pretending that this data-analysis offers any evidence that your conviction is correct. Such a conclusion would be outrageous. Even if we could conclude that Fantunes win IMPs in the bidding (which seems likely, as I wrote above above), it would be almost impossible to distinguish the system wins from the judgement wins, except perhaps by a careful case-to-case study.

To my mind the fact that you present this data as evidence for your conviction makes your claim seem much weaker.

[han]

For what it is worth, I would guess that of the three factors judgement, partnership understanding and bidding system, judgement is by far the most important, and second most important is the level of partnership understandings. Even at the highest level, bridge is still a game of mistakes, and consistently making good choices and having solid agreements is far more important than what those agreements exactly are.

Even if you could find out exactly how many IMPs a pair would win on their bidding, it would therefore be impossible to deduce anything sensible about the quality of their bidding system.

[ulven]

1C revealed. based on same set of data as above.
Total (534 deals): +0.48 imps/board
No overcall (259 deals): -0.13 (!!)
LHO overcall (136 deals): +1,40
RHO overcall (139 deals): +0.71

Breaking down overcalls of 1C:
Simple (240): +1.13
Jump (20): +0.85
Double jump+ (15): +0.07

2-level suit openings lumped together:
Total (540 deals): +0.60
No overcall (246): +0.47
LHO overcall (170): +0.52
RHO overcall (124): +0.98
They overcall/We declare (85): -0.42
They overcall/They declare (209): +1.17

Note: the filter does not state how X of opening bid is classified.
Will check with author.

[ulven]

How would you interprete this?

Uncontested auctions.
Fantoni opens the bidding (646 deals): -0.12 imps/board
Nunes opens the bidding (618 deals) +0.99

There might of course be something wrong with the filter but the author seems to have spent a lot of time to get it right.
Other forum members with collected data might want to double-check.

[han]

ulven, on 2012-August-08, 06:59, said:

How would you interprete this?

Uncontested auctions.
Fantoni opens the bidding (646 deals): -0.12 imps/board
Nunes opens the bidding (618 deals) +0.99

As far as statistical relevance, I don't know what the distribution for the number of IMPs per hand is. I would guess that perhaps the standard deviation is about 6 IMPs per hand (though if it turned out to be 3 or 8 I wouldn't be surprised). In that case the standard deviation for 625 hands is 6/25, or 0.24. That would make the difference more than 4.5 standard deviations, which is a lot (although not enough for people at CERN to conclude anything).

Of course we are dealing with a difference between two tests, rather than a single test. For example, from this data we can conclude more surely that Fantunes scores positively on the hands where Nunes opens, than that Fantunes does better on the hands where Fantoni opens than on hands where Nunes opens. (the former is a single test, the latter is two separate tests.)

Still, it seems likely that this is not a low-data fluke, but that they actually do score better on hands where Nunes opens.

Now perhaps you are asking what we can conclude from that. I'd say: lacking more data we can't conclude anything. Perhaps opener plays more hands, and Nunes plays the cards better. Perhaps Fantoni ends up playing more notrump slams after Nunes opens in a suit, and Fantoni plays those better. Perhaps responder gets to relay for opener's hand, and Fantoni is better at placing the contract. Perhaps Nunes opens more soundly (or more aggressively) and this creates a higher score. Competitive auctions are very different from responder's seat and from opener's seat, but who is doing what different is very hard to judge.

There are so many possibilities, and likely there is more than one factor at play. It is very hard to find out exactly what the cause is. Also, say that you did find out exactly what the reason is, and you see that Fantoni takes better competitive decisions (0.5 IMPs), Nunes plays the hand better (0.3 IMPs) and Nunes opens more disciplined (+0.2 IMPs) and the rest was noise (+0.11 IMPs), it is not clear that that would be interesting to anybody but Fantoni and Nunes.

For a while I kept track of how well my partner and I did on hands where we defended, on hands where he was declaring and on hands where I was declaring. Somehow I thought that this might be interesting. After a while I decided it wasn't. For example, say we have a bad auction to a hopeless slam (sadly, this happened too often). Why is it relevant who is declaring? Perhaps you'd think that these hands even out, but they don't when you play a relay system (which we do).

The only conclusion I drew was that these numbers were fairly meaningless. If you want to decide who does something better, you have to go through the hands manually to see whether somebody made a mistake or did something well. This is very time consuming and incredibly difficult to be objective about. I tried doing this for my own scores, but I know I was not able to be objective.

(I think I did learn something though, just from going through the hands I played very thoroughly and keeping track of what kinds of mistakes we made. Also, often after a couple of days of bridge you feel like you played well or badly. After going through the hands manually, you might find that you made more mistakes than you thought you did, or fewer.)

My answer turned out to be longer than I was planning, and probably less useful than you were hoping for.

[han]

I found in an old bluecalm (thanks bluecalm!) post that the standard deviation is more like 3.5 IMPs per hand (or 0.14 average IMPS for 625 boards) which makes the 1.11 IMPs per board difference even more reliable.

(just a side note: I don't mean to come across as an expert on statistics. I am certainly not, and I could easily be making elementary mistakes.)

[semeai]

han, on 2012-August-08, 09:04, said:

I found in an old bluecalm (thanks bluecalm!) post that the standard deviation is more like 3.5 IMPs per hand (or 0.14 average IMPS for 625 boards) which makes the 1.11 IMPs per board difference even more reliable.

(just a side note: I don't mean to come across as an expert on statistics. I am certainly not, and I could easily be making elementary mistakes.)

Was it this post? It looks like bluecalm is talking about the play-of-the-cards-vs-double-dummy standard deviation as 3.5 imps/board. Later in the same post bluecalm mentions the bidding-(with-double-dummy-play)-vs-par-score standard deviation as 6 imps/board. The par score is e.g. more often doubled than real contracts, so that number is presumably a bit high for more practical concerns.

Your guess of 6 imps/board sounds pretty reasonable, but of course someone with actual data coming along would be best.