riverwalk3

Usla seems set to overtake Leo LaSota within the next 2 months or so as the number 1 masterpoints. One of his strategies from observing his boards is to rebid 3NT after partner responds to 1 of a suit with 1 of another suit, even with a random 13 count and a balanced hand. I'm not sure why how this strategy is effective, but it has to be effective given usla's results.

It worked against humans in a silver lining week tournament: https://tinyurl.com/2hhmfokv

In this deal, my robot opponents bid to a slam off 3 keycards: https://tinyurl.com/25f95txj I knew from the 5NT followup and my own hand that they were probably missing 3 keycards (and thus my partner held an ace), so I doubled them even though they stopped in 6. All 3 keycards took a trick, so they went down 2. Note that East could still have 5 keycards from the bidding (all 5 keycards is 19 HCP, + singleton makes 21 total points, which is stil slightly too weak for opening 2 clubs. [hv=pc=n&s=sj974hk97da9cjt92&w=s62hqj8532d2ckq43&n=sat85hdjt865c8765&e=skq3hat64dkq743ca&d=w&v=n&b=12&a=pp1dp1hp4cp4np5hp5np6dp6hppdppp&p=c5cac2c4d4dad2d5s4s2sas3d6d3d9h5h3c6hthks7s6s5skh6h7h8s8h2c8hah9d7c9hjdjc3c7h4ctsqs9hqstckd8dqcjcqdtdk]399|300[/hv]. However, once East shows the diamond king, West signs off in 6 hearts because it's impossible for East to have all 5 keycards then (that would mean 22 HCP + singleton, which would probably open 2 clubs). If West still thought East had 5 keycards, this would warrant a 7 hearts bid (6 hearts, 3 clubs, 1 spade, 2 diamonds, 1 club ruff).

How would you find such a stopper, when both sides have partial stoppers that combine into a stopper? Does the person with Qx just take a chance (rightsiding the contract if partner has something like Kxx or Axx)?

From my understanding, a robot accepts a claim if there is a sequence of plays that gets at least as many tricks as claimed against any distribution (including ones inconsistent with the bidding). In yesterday's daylong tournament I played, a robot accepted a claim on a double squeeze. This is the link to the hand: https://www.bridgebase.com/tools/handviewer.html?lin=st||pn|riverwalk3,~Mwest,~Mnorth,~Meast|md|2SAQ63HK74DAT8CKT7,S94HT2DKJ97CJ9842,SKT8HAQJ5D3CAQ653,SJ752H9863DQ6542C|sv|o|rh||ah|Board%208|mb|P|mb|1C|an|Minor%20suit%20opening%20--%203+%20!C;%2011-21%20HCP;%2012-22%20total%20points|mb|P|mb|1S|an|One%20over%20one%20--%204+%20!S;%206+%20total%20points|mb|P|mb|2H|an|Opener%20reverse%20--%205+%20!C;%204+%20!H;%203-%20!S;%2021-%20HCP;%2018-22%20total%20points|mb|P|mb|6N|an|4+%20!S;%2017-19%20HCP|mb|P|mb|P|mb|P|pc|C4|pc|C3|pc|D2|pc|C7|pc|S3|pc|S4|pc|SK|pc|S7|pc|S8|pc|S5|pc|SA|pc|S9|pc|SQ|pc|D9|pc|ST|pc|S2|pc|HK|pc|H2|pc|H5|pc|H6|mc|13 At first it looked like an easy 13 tricks, but then East showed out on the first club lead. I won cheaply, then tested spades by leading 3 rounds when the Jack didn't drop, West showing out. I then lead the king of hearts then claimed 13 tricks to see if the robots would accept (not aware of the double squeeze yet), and it did. The robot accepted the claim because there was actually a double squeeze guaranteed once West showed out in spades (implying East was guarding spades), and East showed out in clubs (implying West was guarding clubs): run the hearts, forcing West down to 1 diamond to guard clubs, then cash the top clubs which squeezes East in spades and diamonds (since East discards before you). In fact, I could have claimed as soon as I won the club lead: If West instead guarded spades, then there would be a simple squeeze against West in spades and clubs. If East guarded spades, then there would be a double squeeze. Otherwise, if spades were 3-3, I could win 4 spades without any squeezes. I think the robots should be reprogrammed to accept any claim based on top tricks (with the cards played so far), rather than more complex plays, even if they are 100% lines. Even then though, I often abuse the claim function when I don't remember if a card is good or not and playing that card if not good is risky. Most humans don't accept the claims unless you state clearly how you would play it out, and I doubt any human would accept my claim in that deal unless in an expert match where the claimer clearly specified double squeeze.[hv=pc=n&s=saq63hk74dat8ckt7&w=s94ht2dkj97cj9842&n=skt8haqj5d3caq653&e=sj752h9863dq6542c&d=w&v=0&b=8&a=p1cp1sp2hp6nppp&p=c4c3d2c7s3s4sks7s8s5sas9sqd9sts2hkh2h5h6]399|300|13 tricks claimed, Score = 92.7%[/hv]

Currently, a lot of players earn tops in the Zenith Daylong tournament when they open (or overcall) 4NT nonvulnerable with a weak hand, and go down 6-8 tricks for a top board compared to the opponent's game. This is relatively safe if partner is a passed hand (which when combined with your weak hand indicates that the opponents probably have game). Will this ever be patched? (ie a robot doubles 4NT with a reasonably strong hand). Below is an example. In this case, LHO should clearly double.

Signals are often useful in defense as it gives partner useful information, but sometimes the information isn't too useful for partner and might instead help declarer. For example, if declarer has KQTx opposite Axx, declarer will probably first cash the King and Ace. Assuming the Jack doesn't drop), if declarer knows that the suit is breaking 4-2, then they will finesse the 10 next. In such a situation, it is better to not give an honest count signal to aid declarer. When should you give honest count signals? And how should you play if you don't want to give an honest signal? Just play a random spot card?

There are also a lot of zeroes out of your control: ie your opponent bids a thin game that nobody else does and every finesse is onside, or your opponent stops below game when everything else is in game and suit breaks turn out to be bad. Or your opponent finds a brilliant play that few others find.

I'm mainly thinking of an ACBL regional type event, where you get enormous amounts of masterpoints for getting an overall top (30+ in some cases), and very little otherwise. Additionally, a section top is the only way to make gold points. In this case, it's very hard to imagine someone good enough to consistently get first. Even stars often get scores below 60, from what I'm seeing, when first is usually at least 65. I'm already nearly good enough in robot tournaments to consistently get in the top 3 of my section in ACBL robot duplicates, and am pretty sure Leo LaSota is. In this case a high variance strategy might not be so good (but even then I see Leo LaSota open 1NT with unbalanced 13 counts for example).

I've seen advice to go with the field on the forums. However, isn't higher variance better for tournaments (with the same expected value)? For example, fluctuating between 45 and 65% is better than getting 55% every time, as you will get overall places some of the time with the former but never the latter, and will often get a section top with the former. Thus, it seems better to go against the field, if on expectation the actions are the same. In an extreme example, suppose you have a choice between having a cointoss between 0% and 100% every board, versus getting 50% on every board for sure. In 12 boards, if you choose the cointoss you will get 75% or above 7% of the time, and 67% or above 19% of the time. This means that you will get overall tops very often in this case, while you will barely get any masterpoints at all with the latter.

I have recently started playing 55% Robot Rebates as I know the field is usually very strong there and am trying to challenge myself (I'm not looking to make a profit, which I'm definitely not good enough to, and even players who have won NABC robot individual before such as zhenya_s are only around breakeven). My average probably has been around 52-53% in Robot Rebate (although the sample size is small so this is subject to variance), compared to 60-61% in ACBL Robot Duplicate. While the rebate is certainly harder, I'm not sure it's 8% (I think some of the difference is psychological as I made some weird bids I usually don't in ACBL duplicate). I would be interested in someone else's experience suggesting the difficulty difference. I've also noticed the Zenith Daylongs being much harder (I've broken 70% once as a fluke but often get below 50%), but a closer inspection could give a big part of the difference: I averaged 63% in contracts I declared compared to 47% overall today (and 53% declared yesterday versus 44% overall yesterday).

Suppose South Dealer 1NT P 3NT A, and the Xs are as small as possible.

The scoring is 1 if you get a strictly higher score than the other table (whether by 10 or 1000), 1/2 if you score the same, and 0 if you score strictly below. Hence the finesse would clearly be the normal play without considerations of being ahead/behind the other team.

In Board a Match, say you are in a normal contract and there is a risky play available. Ie suppose you are in 3NT with S xxx H Axx D xxx C AKQx opposite S AKxx H xx D AKJ C xxxx. West leads the King of Hearts. Assuming clubs split 3-2, this gives 9 top tricks. Once you lose the lead, the opponents will run hearts. You can finesse diamonds, but that would risk going down. However, it's probably still the correct play at matchpoints (after running clubs and possible forcing some uncomfortable discards), given that the opponents might only cash 3 more hearts anyway after losing the finesse. The question now becomes: what play should you make if you are currently ahead? Behind? For example, if you are behind, should not finesse, hoping the other team finesses and you get a swing?

Declarer usually outperforms double dummy at the game level, because the opening lead matters a lot and the double dummy defender always makes the correct opening lead. At the slam level, the opening lead matters less, so it is about a wash. At the grand slam level, the double dummy declarer always makes the right play, so the defenders usually outperform double dummy at the grand slam level. Also, in your "bad" examples where North had a lot of points in the red suits, he might bid 4NT instead of 4S, which discourages slam and shows strength in the unbid suits (as 4 clubs was a slam try since it commits the partnership beyond 3NT without assurance of a fit).

Opener has 4 spades (and an off-shape opening) some of the time, so the percentage should increase. Still probably stopping at game narrowly wins out here. It's possible that bidding accuracy could be improved if there was a way for opener to show strength in diamonds/hearts (which would suggest poorly fitting values). Also I wonder how clubs would do in a simulation.

Partner could have duplication in hearts/diamonds, making the singletons worthless. Without duplication, only 30 points is needed for slam (hence why splinters exist; to evaluate fit). 33 points allows for the average amount of duplication, if there is no way to find out. With fancy conventions, partner might be able to show secondary values in hearts/diamond after the 4C rebid, allowing a stop in 4S, but I'd just shoot for the slam if I couldn't find out since according to Richard Pavlicek's methods South's hand is 14 points in clubs and North's hand is 20 points. In your examples, there are tons of wasted honors in diamonds/hearts. But there could just as easily be far less duplication; partner could have AKxx Axx Axx KQx in which you'd want to be in grand slam, and slam can even be made for an ace less after a 1NT opening. The point is to bid slam if there is at least a 50% chance of making. This hand seems admittedly borderline after a 2NT opening, and there is a lot of guesswork involved unless there is a way to find out about the secondary values in hearts/spades (which is much easier after a 1NT opening due to more space).

Richard Pavlicek would count 1 extra HCP for the 4 aces + 10s, and open 2NT with the North hand. Then, after 3H transfer to spades, 3S, 4C, 4S (assuming matchpoints), South's hand reevaluates to 13 points, as 1 extra point is counted for each long club over 3. Thus, 4NT, 5C (0 or 3 keycards) 6S. South's hand also reevaluates to 14 points after a 2NT opening in clubs (2 for the 6th club, 2 each for the singletons, and 2 for the long spade), so the bidding might go 2NT 4C 4N 6C.

I've noticed that in robot ACBL tournaments I can easily get 15 masterpoints per day if I play the whole day, or about 1 per hour. In clubs (live or virtual), the games often last 3 hours, and I struggle to average half a masterpoint per game. Robot tournaments are thus also more cost effective than clubs (since I probably get about half a masterpoint per game, which costs $1.70, or about $3 per masterpoint, while a club game easily costs $5 for only half a masterpoint, or about $10 per masterpoint). Even SYC tournaments (which I average a little over 1 masterpoint per tournament and a tournament takes about 20-25 minutes for me) are more cost efficient than clubs, and much less time consuming to earn the same amount of masterpoints. The BBO daylongs (though not ACBL points) are even more efficient in terms of the amounts of points you get per game; most people with rank 10+ probably got them from daylongs. Doing well on the Zenith Daylong for example earns you 20+ (though the other daylongs also give ridiculous amounts), and if you have a good run you can get 50 masterpoints in 1 day.

Is there any statistic for the percentage of game invitations accepted?

Ouch:

This was the point of the discussion: 4NT natural is an unusual meaning, though possibly it's better to give up the other uses of 4NT in matchpoints when the fit is a minor (in IMPs it means what it used to: Blackwood without other agreements or a substitute cue bid if you play kickback). On each of the hands, you might bypass 3NT but still reach 4NT if you allow its meaning as natural, as the point was that blackwood was often useless with a minor fit due to the lack of space, and stopping at 5 of a minor is usually a poor score. A slow 3NT might just be from lack of stoppers, not necessarily with extra values. Eg in the second example hand I gave S Qx H Axx D KQT9x C Axx opposite your example KJxxx Kx AJxx Jx the top matchpoint spot is probably 4S (a club lead always defeats 6m, and with a club lead 3NT only makes 3 if clubs are 4-4, and it goes down when clubs are 5-3 with the longer club having the As).

This hand is 18 points with a diamond fit (probably 9 cards), 19 by Richard Pavlicek's methods. There admittedly might be duplication in hearts (partner could be short in hearts), but the trump texture is good. The combined strength is very close to 33 points here. Weaken the trumps to KQxxx and probably 3H (which I assume means heart stopper and could right-side 3NT if partner has Qx in clubs) stands out. Possibly a better example could've been S Qx H Axx D KQT9x C Axx which is a perfect fitting hand for partner and only 30 points are needed when the fit is good (Qx in partner's opened suit and prime values in the unbid suits), though I'd actually be nervous about notrump with this type of hand (H/C ace knocked out, then when opponents win the Ace of Diamonds they run their suit), though I could see some players bid 3S in matchpoints here. I wanted QJx of hearts to make stopping in 4NT more attractive.

I tested it using yesterday's results using code. Out of the players who played both the Zenith Tournament and Daylong MP (1) tournament, the average Zenith score was 50.35% and the average Daylong MP (1) Tournament score was 53.13%. The sample size was 373. 2 days ago the Daylong MP (1) average was 52.80% compared to 49.83% in Zenith with a sample size of 432. This suggests that the Zenith Tournament is about 3% harder. The standard deviation for MP (1) is 10.4% while the standard deviation of Zenith is 7.3% out of these overlaps. suggesting that getting a 64.5% in Daylong MP (1) is about as hard as getting a 57.6% in Zenith. Similarly for the players who played both the Zenith Tournament and Daylong IMP (1), the average score was 50.25% for Zenith and 2.22 IMPs for the IMP tournament, with an overlap of 265. The standard deviation for the Zenith tournament was 6.9% and the standard deviation for the IMP tournament was 13.3 IMPs out of the players who played both, meaning that a score of 15.5 IMPs is about as hard as getting 57.1% in Zenith. This suggests that it's abnormal to frequently get 20 IMPs in the Daylong IMP tournament but get in the 40s in Zenith. Out of the players who played both Daylong(1) and Daylong(4), the Daylong (1) average is 50.20% and Daylong(4) average is 51.16%, suggesting that Daylong(4) is ever so slightly easier with a sample size of 191. However, this is probably within the margin of error of chance. The standard deviation for Daylong (4) is 12.39% in these overlaps while the standard deviation of Daylong (1) is 11.27%. For the people who played both the Free (MP) Tournament and the Daylong (1) Tournament on 5/30, the average for Free was 52.50% and the average for Daylong (1) was 48.50%, suggesting that Free is about 4% easier than Daylong (1).

I've noticed that Zenith players always open light - it seems like 11 points is enough for an opening bid in Zenith whereas I usually have 13 in 1/2nd seat (could be weaker in 3rd seat and in 4th seat I open if and only if my hand is above average - 10 HCP with 4 spades is enough for example). I got a bottom for passing a hand with 11 HCP (12 with distribution) and 5 clubs/2 spades in first seat. I think my greatest source of loss is blowing an overtrick while defending however. Slam hands are far less often when you don't have the best hand, but it still seems like slams usually get above average. Even then, if the Robot has the strongest hand and you have the second strongest hand the slam usually doesn't get you that much since the robot is the one pushing for the slam. So the bidding advantage is worth closer to 1/4 as much when you don't have the strongest hand. In best hand slams are usually free points for me since the player is the one making the slam bidding decision. However, when you don't have best hand slam only occurs around 3% of the hands (about 6% of the hands are slams and ), and probably only about 2% of the time are you able to profit (since the Robot puts you in slam). Similarly, when you have the stronger hand you're the one pushing for game. This advantage no longer exists when partner has the stronger hand. Robots don't double as often as they should, so probably competing more aggressively is a winning strategy.