Coin Flip

A place to talk about speculative investing ideas for the optional Variable Portfolio

Moderator: Global Moderator

User avatar
Mark Leavy
Executive Member
Executive Member
Posts: 655
Joined: Thu Mar 01, 2012 10:20 pm
Location: US Citizen, Permanent Traveler

Re: Coin Flip

Post by Mark Leavy » Wed Apr 01, 2020 6:59 pm

Tortoise wrote:
Wed Apr 01, 2020 1:20 pm
Not to discourage learning and experimentation with using the Kelly criterion in options trading, but aren’t any expected gains already baked into the option prices (on average)? Otherwise it seems like hedge funds and big quant groups could swoop in and make a killing using very simple math.
Whatever you do, ppnewbie, don't try this scheme with options. The trading cost, slippage, spread and blowing by your limits would destroy any chance at all of coming out ahead. It's a huge mistake to not count those into your modeling program.

If I were going to give this a shot, I would use some highly liquid, highly volatile ETF that you could trade for free. Maybe one of KBG's triple leveraged Nasdaq ETF's or something like that. One more issue that you would need to factor in: The distribution of wins/losses are guaranteed not to be gaussian. So don't use a random number generator for your simulation. You would be surprised how quickly you can go broke if you assume a normal distribution. Model this with historical tick by tick prices if you can get them.

Should be a fun experiment. If you're careful, it won't be completely stupid and you might do okay.

Keep us updated if you jump in.
User avatar
Mark Leavy
Executive Member
Executive Member
Posts: 655
Joined: Thu Mar 01, 2012 10:20 pm
Location: US Citizen, Permanent Traveler

Re: Coin Flip

Post by Mark Leavy » Wed Apr 01, 2020 8:59 pm

I just did a quick look at TQQQ to see if that would work. It has the volatility you need on a daily basis, but the gaps up and down between the closing and opening prices are huge. You would gap past your exit limits a large portion of the time. Those slips would make this scheme way too risky to consider. Picking up pennies in front of a steam roller.

You would need something that traded round the clock so that you could reliably close each trade at your target percentages.
User avatar
InsuranceGuy
Executive Member
Executive Member
Posts: 279
Joined: Sun Mar 29, 2015 1:44 pm

Re: Coin Flip

Post by InsuranceGuy » Thu Apr 02, 2020 1:11 am

OTM options on SPY/SPX should be able to be calibrated to 50/50.
ppnewbie
Senior Member
Senior Member
Posts: 148
Joined: Fri May 03, 2019 6:04 pm

Re: Coin Flip

Post by ppnewbie » Thu Apr 02, 2020 2:21 pm

One more issue that you would need to factor in: The distribution of wins/losses are guaranteed not to be gaussian. So don't use a random number generator for your simulation. You would be surprised how quickly you can go broke if you assume a normal distribution. Model this with historical tick by tick prices if you can get them.

Good point Mark!
ppnewbie
Senior Member
Senior Member
Posts: 148
Joined: Fri May 03, 2019 6:04 pm

Re: Coin Flip

Post by ppnewbie » Thu Apr 02, 2020 2:23 pm

InsuranceGuy wrote:
Thu Apr 02, 2020 1:11 am
OTM options on SPY/SPX should be able to be calibrated to 50/50.
Thanks for the suggestion.
User avatar
Mark Leavy
Executive Member
Executive Member
Posts: 655
Joined: Thu Mar 01, 2012 10:20 pm
Location: US Citizen, Permanent Traveler

Re: Coin Flip

Post by Mark Leavy » Sun Apr 05, 2020 6:26 pm

Just in case you are still playing with this idea, here's the equations for the mean return, rebalance return (volatility harvesting return), and total return for two assets:. Which is what your scenario is.
(stolen from somewhere on reddit)

I tried this with a handful of numbers representing corner cases and it seemed to come up with the results I would have expected - including the right answer for Shannon's lecture example. So... it looks good on first blush.

Code: Select all

mean return = w1*r1 + w2*r2

rebalance return = w1*w2*[v1*v2*(1-c) + 0.5*(v1-v2)^2]

total return = mean return + rebalance return 

Where,
w1 = allocation to asset 1
w2 = allocation to asset 2 (w1+w2=1)
r1, r2 = expected return of asset 1 and 2
v1, v2 = volatility of asset 1 and 2
c = correlation coefficient

Applying that to your original scenario. (Shannon's example). Each wager is 50% of your net worth and the return is either a 50% loss or a 100% gain. And you make that wager once each day...

This is modeled as one asset being cash (0 expected return, 0 volatility) and the other asset is some speculative, non trending, product that goes up or down randomly and you get out daily at double or half your wager. (0 expected return, 1/sqrt(2) variance).
ShannonsDaemon.png
ShannonsDaemon.png (56.48 KiB) Viewed 307 times
Again, of note, this only works if the returns are somewhat normally distributed. There is no equation for it, but if you run a monte carlo analysis this scheme lags behind buy-and-hold on a product that is trending upward and it loses more than buy-and-hold on a market that is trending downward. So... to make this work better than buy-and-hold (in addition to all of the other real-world issues) you need to use some underlying asset with "random" movements that don't result in any long term trends.

Again, best of luck!
Mark
senecaaa
Full Member
Full Member
Posts: 52
Joined: Fri Nov 29, 2019 4:33 am

Re: Coin Flip

Post by senecaaa » Mon Apr 06, 2020 1:44 pm

Mark Leavy wrote:
Wed Apr 01, 2020 8:59 pm
You would need something that traded round the clock so that you could reliably close each trade at your target percentages.
bitcoin?
User avatar
Mark Leavy
Executive Member
Executive Member
Posts: 655
Joined: Thu Mar 01, 2012 10:20 pm
Location: US Citizen, Permanent Traveler

Re: Coin Flip

Post by Mark Leavy » Mon Apr 06, 2020 1:52 pm

senecaaa wrote:
Mon Apr 06, 2020 1:44 pm
Mark Leavy wrote:
Wed Apr 01, 2020 8:59 pm
You would need something that traded round the clock so that you could reliably close each trade at your target percentages.
bitcoin?
No. The spread is too large.
Futures.
Post Reply