Twister - BRM added! (iPoker 3m)

Hey Tony,

Nice. We look forward to your analysis

Cheers

Laz

Well, as far as strategy, it would be better if winning players analysed this

Starting stacks have 500 chips, each level lasts 2 minutes (but blind increase is smoother, the period of their doubling is 4 minutes), they are:

10/20
15/30 ante 5
20/40 ante 5
30/60 ante 5
40/80 ante 8
50/100 ante 10 [occurs seldom]
60/120 ante 10 [even rarer]
80/160 ante 15 [once in 100 games maybe]

The formal 'rake' (on which awarded player points are based) is 7% of the total BI, so $0.7 for the top $10 level. (In fact it's a bit over $0.71, as the average prize pool is a bit less than $27.87, but the error of this rounding for cashback purposes is relatively small.)

As said, so far the only one of my iPoker skins that has released them is Everest (update: plus Poker770, Betfred and three non-partner rooms so far)... and it's understandable why. The Twister tables are buggy and way heavier on the CPU for some reason:

Originally posted by tonypmm
They are turbo 3-max winner-takes-all SnGs with a randomised prise pool

Honestly, I don't get it. What's randomized if it's winner takes all?

100% of the prize pool are always given to the winner, so there are no deviations in the strategy for a reg (the objective is always to win, no matter if the prize is twice the BI or a thousand times the BI), but for a recreational, there's a fun lottery element - the prize for the first place is chosen randomly.

The idea is much like that of a Golden SnG at Stars.

To put it differently, the total buy-in, say, $10, has three components: $6.67 - the one that forms the guaranteed prize pool ($6.67*3=$20.01~2 BI), the rake ($0.70) and the 'jackpot fee' ($2.63) that's redistributed to rare randomly selected tourneys to boost their prize pools.

Some regs may feel underrolled to play them because it's easy to go on a downswing if one hits a streak of tourneys with the prize of merely ~2 BI (instead of the normal 3).

So Twisters are expected to be quite a soft and also fair format. People won't feel hunted and regs won't engage in lobby gaming like in HUSnG: there's no table selection possible here, it's like 3-max Zoom, but with less frequent reseating, allowing to develop brief reads.

Originally posted by tonypmm

But surprisingly, they've been rolled out not to all skins. In fact, out of the 4 iPoker clients I have installed, only the Everest one has this tab now (above 'My Poker'), and it's very different from the normal SnG lobby

Hey Tony,

Twister will be available on all iPoker skins over the next few days.

Cheers

Laz

^ Which means I have only several days to enjoy the ride until regs from other skins join (It's amazing how recreational Everest's player base is.)

(Well, of course, if the upgrade was delayed for long at other skins, regs would just hop to Everest... or maybe not, it's not like Twister is the game of the century in terms of the holy hourly )

they claim that probability to hit a max jackpot is 1:20k, so if OP is right and it is hyper structured, it should not be a problem to play 20k games?

If it's beatable of course.

You'll need 20K games on average, but playing them by no means guarantess that you hit the jackpot over this sample at all; in fact, the probability of not hitting the max prize during the first k*20000 tourneys (where k is far enough from zero) is close to e^{-k}, where e=2.718281828459045... is the base of natural logarithms, e.g. for k=1 (20000 games) it's 1/e~0.368=36.8%. So the first time of hitting it approximately follows an exponential distribution, and the total number of hits by a certain time (as a function of the latter) is well modelled by a Poisson process.

I was going to write some investigation down in connection with the good old Rio and Fort Knox jackpot SnGs, but sleep issues are chasing me, so please have patience.

Specifically in Twisters, the strategy in a single tournament is independent of its prize pool (apart from tilt issues caused by the money pressure) and of other tourneys - it's always winner-takes-all. That's the key difference between them and the 'old' jackpot SnGs, where adjacent games can influence each other dramatically because a series of six wins is much appreciated (is the condition for winning the JP). And I certainly find the 'old' format more interesting - the frequency of hitting the jackpot is predominantly a matter of skill, not only luck like in Twisters.

On the contrary, outcomes of different Twister tourneys are (almost) independent (just can't influence each other, apart from tilt), so it suffices to estimate the EV of a single Twister (the EV of a series of them is just the EV of a single one times the length of the series). The average prize of a $10 Twister is ~$27.87, iPoker rounds 1/3 of it (the amount that an average player wins back) to $9.30 instead of $9.29 but it's not crucial. The gross win of an average player is the total BI minus the announced fee (e.g. $10 - $0.70 = $9.30), so the average net win is minus the fee. To beat the fee, it's necessary to get cashback and exploit fish, as in any other form of poker. So far the fish seems beatable

I'll think about risk management in Twisters and post the outcome; even though I said that the 'Kelly betting' BRM strategy is the same as if they had a fixed prize pool, it implies going down limits when necessary; some people hate moving down but otoh hate having a risk of ruin over, say, 5%, and a method of calculation of the bankroll needed for a smaller risk while playing at the same limit(s) is well-known and described in 'The Mathematics of Poker' by Bill Chen and Jerrod Ankenman. I'm just too sleepy to plug the numbers right now.

I'm interested to hear discussions about this format. I've won 8 out of 12, including 1 $200 and 1 $40 jackpots.
I haven't seen any decent player in this format yet, only clueless people.
I'm a reg at €10 HU Hypers, so I guess I already have an edge.

Congrats!

Maybe the abundance of clueless people is because skins that have rolled the update out so far (those in whose forums MJPerry has posted and not deleted his 'Twister Poker is here!' threads, i.e. Everest, Poker770, to a lesser extent Betfred) are not very reg-infested. When more regs join and adapt their skill to the format, winrates should go down.

Still, I think it is not very attractive to any HU Sng reg. Afterall, you are probably making more money out of $20 HU Sngs than playing the $10 twister, regardless of the jackpot. Come on, you gotta beat 2 guys to double your buy-in, most of the times, and even when you hit the jackpots, a 3 man hyper turbo has a lot of variance.

So I believe there's always going to be much more fishes than regs on this twister format.
Even after winning this much, I'm skeptical of its profitability, in a way I think my time is better spent playing €10 hu hypers. Maybe it could be more profitable, but then you would need to play so many games it may not be worth it.

But, nevertheless, they are fun ♦

HUSnG regs will switch over not only because the format is close to HU, but because there's no table selection available, and thus there's no table blocking. The idea is much akin to HU Zoom in this respect, with an additional benefit of being able to play with the same opponent for more hands in a row to make brief reads.

Plus I'm really afraid that this range of stakes is only a test one, and the offer will then be extended to higher stakes to rival or maybe substitute old-style jackpot SnGs.

I can see the network's point about removing the latter because recs have little chance in those, as they have low 1st place frequencies, which lowers their chance of grabbing a jackpot by many times, and they don't have good cashback so are eaten alive by the 17% rake. Almost all Fort Knoxes already run without recs.

Still, don't you think 7% rake is kind of a dealbreaker? On HU Sngs it is 4%. So, if we consider the jackpot factor neutral ev, and there are a lot of regs around in twister, wouldn't it be more attractive to stay on the usual HU Sngs, that would have less regs because of twister?

Actually, if you forget about the jackpot, you are playing a $7,37 SNG, and $0,70 is close to 9,5%, 2,4x the rake in a €10 HU Sng

That also depends on your volume - whether you have a real job or grind full time; in the latter case the 7% rake is not a hassle as you'll be able to put in volume huge enough to score well in points races and have 90-100% cashback or (in Rios and Forts) even more (ask NoXz aka Flipp*ndales, the leader of Top100, or Gobias (Ind), or Margar*t1976, the leader of Titan's Grand Battle, who seem to play all of them when they're online), which is harder in HUSnGs.

I've developed the following BRM recommendations for grinding $10 Twisters full time based on the risk of ruin theory from Chen's and Ankenman's 'The Mathematics of Poker':

Win % - pre-cbk ROI (%) - BR size in $

35.0 - -2 - $6550 [regfish]
35.5 - -0.5 - $3450
36.0 - 1.0 - $2550
36.5 - 2.5 - $1500 [leaky reg]
37.0 - 4.0 - $1000
37.5 - 5.5 - $750 [good reg]
38.0 - 7.0 - $600
38.5 - 8.5 - $500 [crusher]
39.0 - 10 - $400
39.5 - 11.5 - $350
40.0 - 13 - $300 [champ]

The assumptions are that:

1. The player has liquid cashback (with essential daily and weekly life expenses deducted) of 45% (e.g. is VIP13 at Betfair or going for it from the very start, so gets instant 50% from point exchange at Betfair with the help of the welcome bonus at first, but has to cash out $250, or 5% of monthly rake, to buy food etc.) and the next bills are due at the beginning of the next month (so he can pay them by race winnings, otherwise should add their value to the said bankroll).

2. The total cashback (incl. points races) minus expenses is 70% (e.g. $4200 minus $700 a month).

3. The player has quite an accurate assessment of the win % (which is never true, lol). If he's a total beginner, then at least extra $100-200 should be allocated for the 'learning curve' period necessary to build a sample of hundreds of games and get the idea of the win %. After this trial period, he should make a very conservative and pessimistic assessment of the win %, and switch to less swingy games if he turns out underrolled. (Of course it's way easier to use such BRM assessments is one is an established player who has just transitioned from another poker variation where he was winning, then he can predict his win % by his historical position in the food chain.)

4. The player can tolerate a 5% risk of ruin. If he's fine with as much as 10%, he can multiply the above bankroll sizes by 0.8; if he can tolerate only 2% risk, he should multiply them by 1.3; if only 1% - then by 1.5.

To compare, that's what I've got for a grinder of €10 Winner-Takes-All Trinities (if they did run at all) who would be only a VIP12 a Betfair and have 50% cashback after life expenses (which eat more because of lower volume), of which 30% total is liquid. (Rake in Trinities is 9.1%, they're of normal speed and with a fixed prize pool.)

Win % - pre-cbk ROI (%) - BR size in $

35.0 - -4.1 - €2100 [regfish]
35.5 - -2.6 - €1550
36.0 - -1.1 - €1300
36.5 - 0.4 - €950 [leaky reg]
37.0 - 1.9 - €800
37.5 - 3.4 - €650 [good reg]
38.0 - 4.9 - €550
38.5 - 6.4 - €500 [crusher]
39.0 - 7.9 - €450
39.5 - 9.4 - €400
40.0 - 10.9 - €350 [champ]

So the variance in the prize size in Twisters indeed makes them a way riskier enterprise, unless one is totally crushing them (which is hard to do given the shortstack turbo structure)...

I'd share the Excel code, but the margin is too thin and I want to sleep; if someone is interested in these intricacies, go ahead and request them...

Your post is golden.
I have 39% of first place with a very small sample (couple hundred games),
but I only have 30% rakeback. I don't withdraw tho.
I'm gonna keep on these until my first deposit bonus expires on poker770, but the idea I see is that twisters need a massive volume, probably more than HU or 6 max Hypers.

You didn't count the VIP program (point exchange), gemstones etc. in, did you? As I've seen you grind twisters hard already :f_wink:, you're not going to have big problems with RB. Select a deal carefully next time and grrrrind it out (a must for any STTs), gl