Will the Tontine surviving player count be 375 (1/19) or less as of 2023-8-20?
27
403
490
resolved Aug 20
Resolved
NO

https://tontine.cash/ is a MSCHF game. On 2021-12-27, 7141 players each paid $10 into a central pot and have been required to check in at least once a day by midnight EST/EDT or else be eliminated. The last remaining player will win the pot. As of 2023-4-2, 486 players remain.

The market resolves YES if the living count is 375 (1/19 of the original population) or less as of 12:05 AM EST on 2023-8-20, and NO if the living count is 376 or more at that time. It resolves early to YES if the living population should fall to or below 376 before that time, and to N/A if the game is canceled for any reason.

Get Ṁ1,000 play money

🏅 Top traders

#NameTotal profit
1Ṁ735
2Ṁ586
3Ṁ438
4Ṁ277
5Ṁ238
Sort by:
predicted NO

382 resolves no

382

LIVING

6.7K

DEAD

601

DAYS

bought Ṁ0 of NO

A day with 4 players dropping out. 9 needed in 4 days. Not seen 9 in 4 days but have seen 8 in 4 days as recently as 23 June. Also 5 in last 2 days is above rate required.

Will this be enough to make this market jittery?

bought Ṁ343 of NO

7 days to go 14 required for a yes resolution(14pw vs 8.5pw required last week). Last 7 days have seen a reduction of 3.

We have seen 12 in a week but not in last 10 weeks

We have seen a week rate about 2.34* what seems a normal rate but even if we call the normal recent rate as much as 5 per week that wouldn't be enough

We have seen 5 multideath days in last 10 weeks and probably getting less frequent. They would have to be large for 4 multideath days this week to do it. A 1 in 14 chance occurring 4 times in 7 tries is very unlikely. It seems too early for someone or some group to pay off 10+ players to drop out but it gets harder and harder to eliminate the last few 0.1% chances.

predicted NO

14 days to go, 17 required for a yes resolution(8.5pw). Last 14 days have seen a reduction of just 5. Just 1 in last week since my 3 weeks to go report seems very low.

Required rate has gone up from 6 per week to 8.5pw.

We have seen 17 in 2 weeks but not for 9 weeks

We have seen an increased 2 week rate but only of about 70% above normal rate and not in the last 9 weeks

I haven't seen a tripling in the 14 day rate in the last 10 weeks let alone 17/5=3.4 * recent rate.

If such an increase were to occur it needs to happen in the next 2 or 3 days or it will need to be noticeably more dramatic increase in rate. Not happened in 70 days and need to happen within 3 days does this get me to less than 1 in 23 chance. It does seem quite possible that more data would get someone to 1 in 50 or lower chance implied by the 2% this claims latest trade.

predicted NO

21 days to go, 18 required for a yes resolution(6pw). Last 21 days have seen a reduction of just 11. 4 in last week since my 4 weeks to go report seems about a normal level.

Required rate has gone up from 22/4=5.5pw to 6 per week.

6pw about 50% higher than recent rates seems unlikely but a high rate of over 1.2 per day for a 21 day period has happened after a longer period averaging under 0.76pd. However I don't know how likely such a high increase in rate is to kick in soon. Last 7 weeks has seen fairly consistent falling rate with the odd bump like down to 13 then up to 16 or down to 10 then up to 12. If it hasn't happened in last 7 weeks is it going to start in next week or two? Seems unlikely. But how unlikely? 1 in 7?

predicted NO

@ChristopherRandles Thanks for keeping this market updated while I sit back and reap the rewards. 😆

bought Ṁ42 of NO

28 days to go, 22 required for a yes resolution. Last 28 days have seen a reduction of just 15 and only 2 in last week since my 5 weeks to go report.

15 in 4 weeks is quite a bit slower than 24 in 5 weeks reported a week ago, but is this a real slowdown or a sampling error that is likely to bounce back to a higher rate? 5 a week to 2 a week seems too large a slowdown and could well be sampling error but 4 weeks is a long period and even if it does go back up to 5 a week that still wouldn't be fast enough from here for a yes resolution.

predicted NO

@ChristopherRandles It's been pretty consistent at 5% per calendar month, so the 15 in 4 weeks is probably just a temporary slowdown. We would naively expect loss of about 19 every 4 weeks.

predicted NO

@jeremiahsamroo Over what period has it been "pretty consistent at 5% per calendar month"?

If you have data going back several months then I don't have the data to argue. We probably should expect a slight decrease in that rate and recent data could suggest it has fallen to ~4.7% per 30 days. 4.7% vs 5% change in direction as expected could be real or just a low period? I don't know without more data, but if you have more data then I am happy to accept that you may well be judging that correctly.

5% of 397 = 19.8

4.7% of 397 = 18.6

18.6 over 30 days reduced to 28 days = 17.4

15 may well be a bit too low to be a sensible expectation but your 19 could also be too high?

or from today

396*.047 *26/30 = 16.1 with the boundary at 20 vs 21

or sticking with your 5%, it is 17.2

Certainly looks like less than a 50% chance of getting down to 375 but the probability may depend on chances of an above average period occurring which probably requires many months of data.

A day with 4 is around 3.3 above average in a single day so it could also happen by chance without a cause for an above average period. So I don't think it is down to a 10% chance at present. Not sure I am any more confident than pick a number between 10% and 40%.

predicted NO

@jeremiahsamroo I doubt it's a temporary slowdown. My intuition tells me that this is permanent.

predicted NO

@ShadowyZephyr Can you put more words to that intuition, or is it just plain mysterious?

predicted NO

@Lorxus I think that the fact that it has settled at 5% was the weird part, when it showed a trend of decreasing before that. Maybe that's based on people who just used scripts having their scripts fail. I believe we've gotten down to the point where the remaining players are going to be much more careful, many are setting alarms too.

predicted NO

@Lorxus We have 396 players who have survived 576 days. An exponential decline i.e. fixed 5% of living players dropping out each month does not seem sensible to me. It should be more like 396 players each with a chance of dropping out averaging at 'some factor' of 1 in 576 days. As it happens the "some factor" seems fairly close to 1. 396/576=0.6875 which is surprisingly close to the current death rate per day. This is the remarkably simple wild projection that I did which suggested it takes 55 years to get down to 10 players.

No I don't believe the 55 years, it is wild to project far ahead, but as time goes on the living players have survived longer so it doesn't seem unreasonable to suggest the game is selecting the more reliable players and that should be modelled as a lower chance of dropping out. Using the number of days survived to model a changing survival rate seems somewhat sensible to me. You seem to suggest you wouldn't believe such a model and at some point the death rate per person would increase. You may be right and 55 years seems silly, but I am not sure how to sensibly model that or explain why some new effect would begin to creep in rather than the longer survival periods continuing to be indicative of more reliable players left with lower rate of dropping out.

predicted NO

@ChristopherRandles i actually mostly agree with that analysis, modulo some quibbling about what that roughly-1 constant actually means. The death rate should indeed be approximately 1/num(days).

However: the quantum of death rate change is much larger in this setting than usual. At some point, it'll be literally impossible to have a survival rate strictly between 0.995 and 1, because there's less than 200 left (f/ex). That's the part I think naive exponential-style predictions likely miss when they predict ends decades out rather than an end in like 2038.

predicted NO

@Lorxus I am not following your 'literally impossible' point. My simple model suggests that if at day 3650 there are 62 players still alive then the death rate per player is 1 in 3650 and there are 62 players so total death rate = 62/3650 so with this I am saying it takes around 3650/62=59 so it takes on average 59 days for 1 player to drop out. We are then at 61 players and day 3709 and so it goes on. Probably totally unrealistic but what else to do instead? Literally impossible - not seeing that.

If bizarrely, the data does more or less follow this projection or something similar, by then it will be obvious the gaps between deaths are getting longer and longer, the remaining players will probably be fed up with it and will try to arrange a settlement.

All projections will be wrong, just not sure if some could be useful.

predicted NO

@ChristopherRandles By literally impossible I mean that if there's 62 players on that day, the survival rate has to be either 1 or something less than 1, and if it's less than 1, it has to be at most 61/62 ~ 0.984 .

predicted NO

@Lorxus We are already at the point where the deaths per day is less than 1. Thus survival rate is already higher than 393/394 and less than 1. So dealing in average days per death doesn't seem like some big thing to me.

bought Ṁ35 of YES

Have you considered making claim(s) in the following manner?

Resolves with PROB=(Tontine living players 31 August 2023 -362)*10

This is capped at 0 and 100 as question cannot resolve to more than 100% or less than 0%.

So 362 or less pays 0%

372 or more pays 100%

Within that range formula in the title applies, so for example if the living player count on 31 August 2023 is 375 then the claim will be resolved at 30%.

predicted NO

@ChristopherRandles I actually hadn't, no. I'd made a numeric market some time ago and set it up poorly, and hadn't really thought to go back to that or YES/NO markets which, by resolving to a PROB, basically do the same thing. I think I agree that it might be worth doing that again, especially in the case you mention where it's way finer-grained.

predicted NO

@ChristopherRandles Also now that I remember as much, I don't actually get that many traders on these markets, and I feel like if I tried to do something that complicated, I'd get none.

predicted YES

I've lost mana on every single tontine market I have partecipated in. I wish there was a prize for the worst tontine predictor.

predicted NO

@Odoacre I mean, I could always make one, given that all of these markets are in a group, and that group has a leaderboard...

bought Ṁ30 of NO

5 weeks 24 required. Last 35 days saw 24 reduction so at same rate it happens but 10 week period is time for the rate to slow down as there are fewer people to potentially fail and the remaining people get more reliable. 50% chance, maybe a little under, unless there is good reason to think last 5 weeks was unusually slow?

9 day period with only 2 deaths did look unusually slow but that is likely cherry picking and out of 5 weeks the effect does not seem all that huge.

bought Ṁ25 of NO

@ChristopherRandles 23 required, but yeah, same idea. For some reason it took a while for someone to point this out and I was able to accumulate a large amount of NO shares.

predicted NO

My credence is like 35% if you're curious

predicted NO

@ShadowyZephyr Fascinating! Mine's about 70%.

predicted NO

@Lorxus Did it now change? You bet no down to like 40%

predicted NO

@ShadowyZephyr Very much so yeah. I have to manually update some things in my sheet - I did and it crashed to like 45%. I agree with you now more or less.

predicted NO

@Lorxus Mission accomplished :D