Peculiar hybrid format: Any strategy tips?

I've qualified online for a live tournament. (Qualification just involved meeting a wagering requirement.) Having previously been to Bulgaria, I'm getting another free trip, this time to Prague in the Czech Republic.

I've just been sent the rules, and they look a little strange -

There are 18 players, with 6 getting paid.

There will be semis and then a final table, but the semis and the final will each consist of 3 separate rounds of 21 hands. A round is conducted a bit like a table-elimination event (i.e. with a rotating dealer button). But it's really an accumulation event - at the end of a round your chip total goes onto the leaderboard (and your scores for the three rounds are ultimately added up to give your overall score). The top 6 overall scores from the semis go to the final table, where the process is repeated from scratch to determine the final, top 6 positions.

The betting limits are min:100 max:1000, with a starting stack for each round of 5000.

I presume the semis will be run as 3 tables of 6. There is a seat draw prior to each round (presumably meaning you needn't find yourself with the same 5 other players in each round of the semis).

I guess a big factor in deciding how to play the semis will be whether or not it proves to be possible to monitor what is happening at the other tables other than by noting the scores after the each round.

As far as I can see, the final, final table will end up being a variation of the usual table-elimination-style final table, where you not only need to try and keep track of the other players' chips in front of them, but also remember what their scores are from the previous two rounds. Sounds horrendous!

Any thoughts about how to tackle this?

 

I've played a a few events with a similar format. In these cases, the first round was broken up into two sessions of 21 hands each with the intermediate scores posed after the first session. Pretty much the same as what you describe, except that it's 2 sessions instead of 3.

I've had pretty good success advancing at these events. For me, the most important thing has been determining what the threshold bankroll for advancing will be. Wong's formula is ok for tournaments with a large number of players, but it will likely not be of much help here.

Here are my methods for determining the goal:

• Ask a few tournament officials what the threshold has been in the past. They can usually tell you the highest, lowest and typical thresholds that they have seen.
• Get friendly with some of the locals. Without revealing that you're trying to establish a goal, ask them in general how they approach the tournament or, even, better, tell them that you have no idea how to approach it. They will either reveal that they have no strategy (they think it's all luck) or they will tell you exactly what their goal is. Most players seem very eager to educate someone whom they determine is less knowlegable than themselves. Pretending to be a newbie is very effective here. Usually a consensus among the experienced players emerges.
• Experience. I keep records of the advancing thresholds I have observed at each event.

In the absence of any of the above (one-time or first-time event) you can use Wong's formula for large tournaments. For a small one like this, you will likely be able to pay some attention to the other tables during your sessions.

The fact that there are multiple sessions is very valuable. For the two session tournaments I've played, I've found that doubling the BR of the threshold player after the first session has proven to be a very accurate estimate of the final threshold. Perhaps tripling it in your case after the first session and then re-evaluating it again after the second session would be a good approach.

Whichever method you use to establish the threshold, set your goal slightly higher as a margin for error. This is especialy effective against the players who have told you their own goals. For example, in one tournament I play, the consensus goal among the experienced players is to reach $20,000 and, in fact, $20,000 is pretty close to the average actual threshold. In the final results, there always ends up being a large cluster of players within $250 of that $20,000. By setting my goal to $21,000, I end up ahead of a large number of players who are near the actual threshold.

Once you have determined your own goal, I believe that the best strategy is to bet max until you achieve it. Variance is your friend here, so take all double down and split opportunities. You might even consider a few borderline cases, like doubling A,7 vs 2. If time (number of hands) is running out, you should, of course, get very aggressive with your doubling and splitting. If you get close, and the betting limits allow it, then you can use a progression to improve the odds of making it. If it is the final hand, and you have not made your goal yet, then bet max. Once you have reached your goal, bet minimum. At this point, I try to reduce variance by not doubling or splitting and I use the playing strategy for "a push is as good as a win".

One way to think about it is that you are chasing an opponent who is sitting out.

One other thing. Since, in this tournament, you get a fresh bankroll in each session, don't forget to include the bankroll for upcoming sessions in your running total. For example, say you have determined that your final goal is $25,000. At the start of each session you are given $5,000, so you only need to win $10,000 (over 3 sessions) to reach your goal.

Another example. Say that you won $6000 in the first session (for a total of $11,000) and that at some point during the second session you have won another $5000 (for a total of $10,000). You have now reached your goal, since you have a running total of $11,000 + $10,000 plus the $5,000 you will be given to start the third session.

 

It might not be as bad as this. I've always found that the lead board is within sight of the tables at these events.

I would approach the final table like a single 63 hand round with a $15,000 bank roll. The only real differences are that if you lose $5,000 in an early session, you have to sit out a few hands until you have access to the rest of your bankroll and that, in the final session you only have $5,000 to work with. With the $1,000 max (very small compared to the overall bankroll), I'm torn as to whether to be very aggressive or conservative.

If you're aggressive (I would consider betting max in the first session) you have a decent chance of building up a 2+ max bet lead, but still have plenty of time and ammunition in the second and third sessions to play catch up should be unlucky.

If you're conservative, you can enter the final session with somewhere near a $15,000 running total and are likely to be within range of the leader(s) with 21 hands to play.

My software has a module which can simulate different tournament formats using bots of varying levels of skill and aggression. It has proven useful in setting my own agrression level for various tournaments I play in. If you can give me the exact playing conditions and rules for this tournament, I can see what it comes up with.

 

Thanks, Gronbog. That seems like a lot of useful advice.

But part of what I was wondering about is whether the fact that an elimination-style rotating dealer button will be in use makes much difference. Bearing in mind that in each of the first three rounds I will be seated at a table consisting of a third of the total field, then using elimination-style strategies to aim to finish my table in the top two or three might be a worthwhile tactic (at least until it becomes clear that a number of players at other tables and/or in previous rounds have built up a big lead).

Certainly in the third round, if I happen to know that my biggest rivals have been drawn to the same table as me then there is some scope to make use of that knowledge.

I've not played in any accumulation events before. I presume they don't normally bother with a dealer button?

 

Yes, I think that is likely. But even then I'd have to remember who is who, plus do the mental arithmetic of adding their previous scores to their current chip stacks.

I've played few live events and not really mastered the art of chip counting. So these added headaches would be a bit much for me. Still, first I need to qualify, so there's every chance it won't be an issue!

I normally tend towards the conservative, but I'm not sure what's best here, given all the factors at work.

I think perhaps being contrarian from the outset might be a good approach. If there are lots of big bettors I can play my usual game, but if not I need not wait as long as usual before making a move.

Thanks very much.

I'm afraid I don't know the exact playing conditions, beyond what I've described. I've asked for a bit more information, but there's no guarantee I will get it before the event. (For instance I don't know the BJ rules in the casino. If it should be the case that you can't double on any two cards, that would make a big difference.)

 

In a large tournament, the puck and what is happenning at other tables is largely irrelevant. In a small tournament, like this one, it can be relevant if you can see the chip stacks on the other tables. In this case, where 6 of 18 advance, shooting for the top 2 at your table might not be a good idea because if one other table is hot, then that entire table could advance over you.

Certainly there is some useful information here, especially if you see that more than 6 other players have exceeded your goal and are coasting.

Sometimes they do, sometimes they don't. Some superstitous players (and tournament officials) don't think it's fair unless everyone has to be the anchor equally.

 

Now this is something I can help you master. I was an online-only player, too, until my very first live BJT table.
A UBT final table taped for TV in Hollywood, no less, and I had to learn chip counting in a seriously big hurry.

Get a set of good quality poker chips (I borrowed a friend's case of WPT chips). Hopefully there's someone else in your household, get together with them and agree on a value for each color. Have them "hide" stacks of chips in various places in your home. Some of the hiding places I found were:
On top of bedroom dresser
Next to the coffeepot
On the dashboard of your car
In the refrigerator (that one got me laughing when I found it)
Next to the bathroom sink

You get the idea. The rules are that you have to stop what you're doing and quickly count the value of the stack immediately by sight alone and from 4 to 6 feet away. Once you've got an answer, then you can grab the chips and count 'em up. Not only do you have to give the answer but you must also give the answer for the previous stack, too. This helps get your brain thinking about counting stacks and remembering results even when you might not be at your best.

 

Thanks

Sounds like every day is Easter in your house, Lefty.

 

I got some answers regarding the rules -

No Surrender.
No Doubling for less.
And only double on 8,9,10,11. (So I guess a lead of > 1 max bet would be hugely significant)

 

OK. I'll try simulating the format this weekend and see how the various bots do.

Once you've actually played the tournament and observe the actual strategies used by the other players, we can do even better by limiting most of the bots to those styles. Of course, this will only be a useful exercise should get the chance to play again in the future.

 

Thanks again, Gronbog.

For what it's worth, I think it will also probably be a ENHC game, with 6 decks.

It sounds like there may be future events, but I'm not sure how frequent. I'll find out all I can when I'm there and post the details.

 

Do you know the prize structure? With this information, I can tell you which bots do best in the final round with respect to expected prize money vs trying to win the tournament outright.

 

The six finalists get paid as follows -

1st: 8000, 2nd: 5000, 3rd: 4000, 4th: 3000, 5th: 2000, 6th: 1000

How do your bots interact with one another? Do they look at the BR totals after each hand, or just play their own strategies in isolation? (I'm assuming they don't take notice of opponents' bet sizes on a specific hand for correlation/contrarian betting purposes?)

I guess there are two separate questions here - the best strategy to qualify for the final table, and then the best strategy to employ once there.

 

Thanks! I'll incorporate this into the sims.

The answer is all of the above. Some of the bots are stupid and apply some simple strategy in isolation. Some look at the BRs and bets of the other players. Some take this further and leverage their button position, including correclation of bets. Some use varying levels of various strategies I've read about or that I've devised myself. There is even a bot that is intended to play like me (very difficult to program and still not perfect).

The main thing is that I have actually played against folks, either online or for real, who use each of the strategies that I've programmed (or some variation). Once I've played an event a few times, I can limit the simulation of that event to mostly bots who play the strategies that I've observed and in the rough percentages that I've observed them. I leave room in each tournament iteration for all of my bots to randomly play in and see which ones tend to to the best.

Yes. The data is broken down into each round of the tournament, so I can see which strategies tend to do better in each separate round and also which bots tend to win the most by switching strategies to suit each round.

 

More information. Can you tell me what the betting unit is within the 100 min, 1000 max? i.e. can one bet 105? 110? 125? or must it be in units of 100?

 

That's something I forgot to ask, but I'm working under the assumption that bets will have to be multiples of 100. That seems the most likely.

 

Sim Results

Sorry this took longer than I had thought. I had to do a bit of programming to allow modeling the cumulative style final table sessions and I wanted to make sure that there were no bugs before I posted anything.

I have some results which I hope will be of some value to you. Let's start with the cumulative preliminary sessions. The model is as follows:

Rules:
6 decks, 83% Penetration (5 decks), ENHC, double 8, 9, 10, 11 only, no double for less, split up to 4 hands except aces (1 card each), no surrender, BJ pays 3/2, no doubling on BJ.

Format:
18 players, 3 sessions of 21 hands each. Starting bankroll of 5,000 for each session. Max 1,000. Min 100. Betting in increments of 100. Top 6 after the 3rd session advance to the final round. The final round is the same again with 6 players all at the same table.

Assumptions:
There was some discussion about what information would be available regarding the action at other tables. I made the following assumptions based on my own experience of playing at cumulative events and what I believe you will be able to find out should you be motivated enough:

1. The leaderboard from the previous session will be available to you either because it is directly visible from your table or because you were able to write it down before your session began.
   
2. Because it is such a small event, you should be able to determine who is at your table. This can be done by mingling between rounds and learning people's names or by introducing yourself at your own table and hopefully getting names in return. First names should generally be sufficient. This assumption applies to both the preliminary sessions and the final table sessions. This assumption is particulary important in the 3rd preliminary session and is vital during the final table sessions.

If you can see what's going on at the other tables during the preliminary sessions (especially the 3rd one), then even better.

About the Bots:
Here are some things to keep in mind about the bots. They all have some finishing skills which are used in the last 5 hands for catching up or for protecting a lead. Other than that, they all employ some fundamental philosophy for playing the rest of each round. For example, bet the table average or bet contrary to the leaders. Some of them use very simple strategies like bet max until you're ahead. Some of the fundamental strategies are agressive and some are not. That's what we're mainly interested in.

With this in mind, regardless of which bots perform the best in a given tournament format, we, as humans, can keep the basic philosophies of those bots in mind and do a little better ourselves by applying variations to those strategies as appropriate.

Preliminary round results:

The following table shows the goals which achieved various probabilities of advancing during the simultation. For example, achieving the goal of 18,900 was good enough to advance 70% of the time. To guarantee advancement, one would have had to achieve 31,250. Note also, that the mean advancing threshold was ~17,640, which means that a passive strategy is not recommended.

Leader Board Statistics
************************************************
Number of samples == 442085
Mean Advancing Threshold == 17640.099868
0% Quantile Advancing Threshold == 6000.000000
10% Quantile Advancing Threshold == 14350.000000
20% Quantile Advancing Threshold == 15300.000000
30% Quantile Advancing Threshold == 16000.000000
40% Quantile Advancing Threshold == 16650.000000
50% Quantile Advancing Threshold == 17350.000000
60% Quantile Advancing Threshold == 18000.000000
70% Quantile Advancing Threshold == 18900.000000
80% Quantile Advancing Threshold == 19900.000000
90% Quantile Advancing Threshold == 21450.000000
100% Quantile Advancing Threshold == 31250.000000
************************************************

In fact, the bots with the best advancement rates were all bots which bet maximum until they are in an advancing position on the adjusted leader board (posted leader board adjusted by the scores at their table) and bet minimum otherwise. The most successful bot ("Simulated Gronbog") was able to advance 41.15% of the time

The least successful bots were ones which tended to bet low, including all of the bots which bet 35% of the maximum or less. The least successful bot ("Elim Average" -- bets the average of the leaders' bets in an attempt to keep up) was able to advance only 20.31% of the time.

Final Table Results:

Winning the Tournament:

For winning the tournament, once in the the final table sessions, the most successful bots were, once again, the aggressive ones. Most of the top 10 bots were ones which bet maximum (or close to it) when not first and minimum otherwise. The most successful bot as "Full Maximum" which simply bets max all of the time. This bot won the final table round 24.47% of the times that it made it.

Once again, the least successful bots were ones which tended to bet low, including all of the bots which bet 25% of the maximum or less. The least successful bot was "Spectrum 5%" which flat bets 5% of the maximum (10% in this case due to the 100 betting unit) for a success rate of only 6.1%

Maximizing Prize Money:

For maximizing prize money, once in the final table sessions, the most successful bots were, those which bet contrary to or which chased the table leader (ignoring all other players). The most successful was "Chase Leader" which bets enough to catch the leader, or max if more than a max bet behind. This bot averaged $4110.51 each time it made the final round.

Once again, the least successful bots were ones which tended to bet low, including all of the bots which bet 35% of the maximum or less. The least successful bot was "Spectrum 5%" which averaged only $3568.52 each time it made the final round.

Best Overall Bots:

These are the bots which had the best combined stratgies for all rounds.

For Winning the Tournament:

Once again aggression wins out. The bots with the best chance of winning the tournament from the outset were the ones which bex maximum or close to it until they were on the adjusted leadboard during the preliminary sessions and which bet big when behind during the final table sessions. However, the most successful bot overall was "Full Maximum" (bets max all the time) with a 8.66% (1/11.54) chance of winning the tournament from the outset.

Once again, betting small was a formula for disaster. The least successful bot was (also once again) "Spectrum 5%" with on a 1.25% chance of winning the tournament from the outset or 1/80.

Maximizing Prize Money:

For maximizing prize money from the outset, the best bots were the ones which bet big in order to make the adjusted leader board during the preliminary sessions and which bet contrary to the leader at the final table. The best overall bot was "Simulated Gronbog" with an expectation of $1596.77 each time it played the tournament.

This bot starts with a goal of $30,000 in the first preliminary session and bets max in an attempt to reach it. After the first session, the goal is adjusted to 3 times the BR of the threshold player. After the 2nd session, the goal is re-adjusted to 1.5 times the BR of the threshold player. For the final table rounds, this bot bets contrary to the table leader in general, and will aggressively chase if the leader gets a couple of max bets ahead.

You guessed it. The small bettors lose out again.

One thing worth noting is that even the worst bot ("minimum") managed to average $729.54 in prize money each time it played, making this a profitable event even for the worst players.

One final note. I have a bot which bets randomly within the table limits with no regard to anything else. Even this bot managed to average $1,170.35 in prize money playing this event.

So maybe the best advice of all is to not worry too much and have fun!

 

Doubling only 8, 9, 10, 11

This comment got me thinking about how the percentages for the common situation of needing to win two bets would change with these rules and what the optimal strategy would be. It's fairly obvious that we must double or split our initial hand, but which pairs should be split and which should be doubled with these rules? Also, how should we play our hands once split?

It turns out that the optimal strategy is to double when you can (4,4 and 5,5). We are forced to split otherwise.

Once you have split, the optimal strategy for the first hand is also fairly straight forward:
- double if you can
- resplit if you can
- stand on all hard hands 12 or higher (no bust policy)
- stand on soft 18 or higher
There are a few exceptions, but the percentages are extremely close and not worth worrying about.

After the first split hand, there are too many combinations to list but a no-bust policy would likely go a long way.

The simulation result for the optimal strategy was a 12.05% success rate which is much less than the generally accepted success rate of ~30% for DOA and very close to the ~12% rate for a win/lose swing. It would seem that any lead which allows you to cover your opponent's single bet win would be extremely valuable on the final hand and that Curt's Revenge would not be a viable option.

 

Thanks for your efforts. I've printed out your results to give me something to read on the plane.

I'll certainly try to follow that advice.

I fly out later today. Things have got off to an inauspicious start - I had also qualfied for a big online final at blackjack21 which took place at 3:40am today, meaning I've been up since about 3am as I write this. I had to schedule my travel plans around it, but needn't have bothered since I was knocked out in the first round!

At least if I have similar luck in the live tournament there'll be some tourism for me to do, to take my mind off it. (Plus a party, apparently.)

 

I wonder how that works out if you are BR1 with, say, all five opponents within range to overhaul you if they win a double.

If you want to maximise your chance of staying BR1, I don't know if it would be better to risk the swing by covering them all (as you would if it were DOA), or bet small and hope that they all either can't double or lose their doubles.

The answer might also depend on another question I haven't yet asked them - Can you split unlike tens? If you can't, then I imagine that significantly reduces the chances of winning 2 bets on a random hand.