What's the Risk?

General TSP Discussion.

Moderator: Aitrus

User avatar
evilanne
Posts: 2067
Joined: Thu May 14, 2015 6:52 pm

Re: What's the Risk?

Post by evilanne »

So how are you translating all of this to your selection of a daily strategy. Once you have 3 standard deviation or greater, what is the point? After looking at a bunch of the different mixes, they seem to be very close to each other...substituting G for F or changing a day here or there. If that unexpected one in a million (or whatever measure you use) chance happens at the wrong time, I don't think the standard deviation is going to matter if you are in the market.

User avatar
fordest
Posts: 605
Joined: Wed Sep 26, 2012 3:48 pm

Re: What's the Risk?

Post by fordest »

evilanne wrote:So how are you translating all of this to your selection of a daily strategy. Once you have 3 standard deviation or greater, what is the point? After looking at a bunch of the different mixes, they seem to be very close to each other...substituting G for F or changing a day here or there. If that unexpected one in a million (or whatever measure you use) chance happens at the wrong time, I don't think the standard deviation is going to matter if you are in the market.


That's a good point. It's a similar argument we were making when people were trying to get a strategy with 38.32% as opposed to 38.12%. It really wouldn't make a difference because either one would have similar chances to succeed or fail.

I think it's just become a challenge now maybe . :lol:
100% in the daily system since August 2, 2017.
Following strategy (current pick) #88676. 2020 real life has been following #110838

User avatar
12squared
Posts: 675
Joined: Thu Dec 31, 2015 7:28 am

Re: What's the Risk?

Post by 12squared »

evilanne wrote:So how are you translating all of this to your selection of a daily strategy. Once you have 3 standard deviation or greater, what is the point? After looking at a bunch of the different mixes, they seem to be very close to each other...substituting G for F or changing a day here or there. If that unexpected one in a million (or whatever measure you use) chance happens at the wrong time, I don't think the standard deviation is going to matter if you are in the market.

Similarities among the high sigma patterns suggest that there actually are repeatable patterns underlying market dynamics. They tell us that "when" you are in the market CSI matters more than how much time you spend there.

Even for strategies with standard deviations as low as 6 the margin of error (in the mean) is +/- 3 percentage points (with 95% confidence). At an SD of 20, it is +/- 10.5 points. This implies that all of the 100 strategies in a TSPCalc list sorted by Mean are in a statistical dead heat. If you use Mean as your gauge, one is as good of a choice as any of the others. On the other hand, small differences in the sigma factor have a huge effect on uncertainty, and its cousin, reliability. Ignoring margins of error for the moment, 2σ is 7x more reliable than 1σ, 3σ is 17x more reliable than 2σ, 4σ is 42x more reliable than 3σ, 5σ is 110x more reliable than 4σ, and 6σ is 290x more reliable than 5σ.

Given a choice of 2 vehicles at the same price, style, trim, etc., wouldn't it make sense to choose the one known to be more reliable? Is taking a 1 in million chance on a lottery ticket a good bet? If not, 1 in a million odds against something happening should seem exceptionally good.

Any 5σ strategy with a mean of 30+ will let me sleep just fine. The biggest risk may in not having access to TSP.gov during the 24 hours before an IFT is on the schedule.
“The genius of investing is recognizing the direction of the trend – not catching the highs or the lows.”
- Dean Witter

"Put all your eggs in one basket and then watch that basket."
- Andrew Carnegie

User avatar
Aitrus
Moderator
Posts: 2391
Joined: Mon Aug 06, 2012 5:03 pm

Re: What's the Risk?

Post by Aitrus »

12squared wrote:Similarities among the high sigma patterns suggest that there actually are repeatable patterns underlying market dynamics. They tell us that "when" you are in the market CSI matters more than how much time you spend there.


This applies to the monthly seasonals as well, it would seem. All of the best performing monthly strategies are in the market from Feb / Mar - May, are out in June, in for Jul, out for Aug/Sep, and in for Oct - Dec. The main differences between the Mixes are the monthly Fund selections long the way, but the pattern of in/out is largely the same.
Seasonal Musings 2022: viewtopic.php?f=14&t=19005
Recommended Reading: http://tspcenter.com/forums/viewtopic.php?f=14&t=13474
"It's not what happens to you, but how you react to it that matters" Epictetus

mindofmush
Posts: 353
Joined: Mon Jul 02, 2012 1:38 pm

Re: What's the Risk?

Post by mindofmush »

That's a lot of nifty numbers but I'm from Missouri; I need you to show me how a 5σ strategy is 78,540x more reliable than a 2σ strategy. Will the 5σ strategy have positive returns for each day in it's designated fund?
If we ran these strategies side by side for comparison, how many days/weeks/years will it take to demonstrate a significant superiority?
Can I use #15823 (a 6σ strategy) and rely on 30% return (and 5+ million) over the next ten years?
mo meng, mo ching (which loosely means: no money, no life)

User avatar
12squared
Posts: 675
Joined: Thu Dec 31, 2015 7:28 am

Re: What's the Risk?

Post by 12squared »

mindofmush wrote:That's a lot of nifty numbers but I'm from Missouri; I need you to show me how a 5σ strategy is 78,540x more reliable than a 2σ strategy.

I couldn't tell you were from the Show-Me state by your flag - but I'll give you the benefit of the doubt.

The term "sigma" is used in the world of Quality Control as a metric of reliability. It is akin to the use of the decibel (dB) for sound and radio signal levels, in which ratios span great ranges, e.g. from micro to giga. Reliability is usually expressed as a 1 in a______ chance of defects. In the context of annual returns, I defined the defect limit as the 0% return, because anything less results in "loss of capital".
Image

The relative reliability of sigmas was calculated by dividing their respective Cumulative Normal Distributions obtained from Excel's NORMDIST function.
Image

mindofmush wrote:Will the 5σ strategy have positive returns for each day in it's designated fund?

Because all the statistics have been run on yearly data sets, they should only be judged on an annual basis. For those who need this level of certainty on a day to day basis, there is always the G fund, which some claim is "safe as kittens".

mindofmush wrote:If we ran these strategies side by side for comparison, how many days/weeks/years will it take to demonstrate a significant superiority?

Probably about as many cycles as it takes to determine the "best" route to take to work - under all weather and traffic conditions.

mindofmush wrote:Can I use #15823 (a 6σ strategy) and rely on 30% return (and 5+ million) over the next ten years?
Any probabilities or reliabilities determined using the historical data are best used for relative comparison. You can use whatever strategy makes you most comfortable. Nothing is 100% certain - except that G Fund matches inflation. However, your chances having a positive return are 2 orders of magnitude better with something whose combination of Mean and StdDev equates to 6σ than to 5σ.
“The genius of investing is recognizing the direction of the trend – not catching the highs or the lows.”
- Dean Witter

"Put all your eggs in one basket and then watch that basket."
- Andrew Carnegie

User avatar
evilanne
Posts: 2067
Joined: Thu May 14, 2015 6:52 pm

Re: What's the Risk?

Post by evilanne »

Looks like onerepmax came up with quite a few more 7 sigmas, 15982 has the highest RoR so far for standard deviations under 5 but doesn't quite measure up to a 7, but that could change with more data.

http://tspcalc.com/seasonal.php?ID=1598 ... sc&SDmax=5

User avatar
fordest
Posts: 605
Joined: Wed Sep 26, 2012 3:48 pm

Re: What's the Risk?

Post by fordest »

I go away for the weekend and come back to some 7's ... Nice work everybody!
100% in the daily system since August 2, 2017.
Following strategy (current pick) #88676. 2020 real life has been following #110838

User avatar
12squared
Posts: 675
Joined: Thu Dec 31, 2015 7:28 am

Re: What's the Risk?

Post by 12squared »

evilanne wrote:Once you have 3 standard deviation or greater, what is the point?

The higher the sigma, the greater the chance of a positive return. A strategy in which 0% is 6σ below the mean (i.e. Mean/StdDev = 6) will have a 2σ chance of a return at least 4 StdDevs above 0%. Or a 3σ chance of a return at least 3 StdDevs above 0%.
“The genius of investing is recognizing the direction of the trend – not catching the highs or the lows.”
- Dean Witter

"Put all your eggs in one basket and then watch that basket."
- Andrew Carnegie

mindofmush
Posts: 353
Joined: Mon Jul 02, 2012 1:38 pm

Re: What's the Risk?

Post by mindofmush »

I'm sold.
Wait, getting a nagging feeling,
The numbers look reassuring and the STD DEV are really flat BUT the tspcalc uses over 13 years of daily data (over 2500 data points for each fund) which sounds like a large enough sample size, but we're not analyzing the database, we're only looking at the annual results, only 13 data points. I thought a sample size of 30 was considered the smallest to be representative of the whole.
I believe that for a sample this small that evilanne is right, that more than 3 sigma will not improve the outcome.
mo meng, mo ching (which loosely means: no money, no life)

User avatar
onerepmax
Posts: 138
Joined: Tue Aug 08, 2017 9:29 am

Re: What's the Risk?

Post by onerepmax »

Here's slew of strategies with well over 6 mean/SD ratios for those who are a bit shy on the F fund. All under 15% time in F. Two at 9, and one at 10.

Strat # Mean SD ratio
16024 30.66% 4.70% 6.52
16090 30.45% 4.70% 6.48
16020 30.00% 4.42% 6.79
16018 30.15% 4.32% 6.98
16015 30.16% 4.25% 7.10
16017 30.14% 4.34% 6.94
16016 30.12% 4.25% 7.09
16014 30.28% 4.29% 7.06
16013 30.13% 4.16% 7.24
16071 30.88% 4.45% 6.94
16010 30.36% 4.40% 6.90
16011 30.35% 4.23% 7.17
16012 30.23% 4.24% 7.13
16008 30.47% 4.47% 6.82
16009 30.39% 4.38% 6.94



http://tspcalc.com/seasonal.php?ID=1609 ... 75&Fmax=15

I don't know the exact correlation between the ratio and the sigma number... but bigger ratio is better.

User avatar
12squared
Posts: 675
Joined: Thu Dec 31, 2015 7:28 am

Re: What's the Risk?

Post by 12squared »

onerepmax wrote:I don't know the exact correlation between the ratio and the sigma number... but bigger ratio is better.

The Mean/StdDev ratio is the sigma multiple that gets you a 0% return.
“The genius of investing is recognizing the direction of the trend – not catching the highs or the lows.”
- Dean Witter

"Put all your eggs in one basket and then watch that basket."
- Andrew Carnegie

Regularguy
Posts: 191
Joined: Tue Sep 24, 2013 11:39 am

Re: What's the Risk?

Post by Regularguy »

ksmoly04 wrote:This is great information 12x12. This puts alot of the daily strategies into much better context. For those interested:

7980: ratio 2.2, guaranteed negative return 1.39% of time; rarely more than 9.9% of time.

What this means is a potential delay in my hitting the $1M mark by 10 months (when using the aforementioned Strategy #15204. Theoretical tsp amounts at MRA retirement (2038) are cut in half.



Is that 1.39% of the trading days, or 1.39% of the 28 trades will be negative?

User avatar
12squared
Posts: 675
Joined: Thu Dec 31, 2015 7:28 am

Re: What's the Risk?

Post by 12squared »

Regularguy wrote:
ksmoly04 wrote:This is great information 12x12. This puts alot of the daily strategies into much better context. For those interested:

7980: ratio 2.2, guaranteed negative return 1.39% of time; rarely more than 9.9% of time.

What this means is a potential delay in my hitting the $1M mark by 10 months (when using the aforementioned Strategy #15204. Theoretical tsp amounts at MRA retirement (2038) are cut in half.



Is that 1.39% of the trading days, or 1.39% of the 28 trades will be negative?


Because the data is on a yearly basis, all probabilities are also on a yearly basis. The Mean/Std for each month is much lower than on an annual basis
“The genius of investing is recognizing the direction of the trend – not catching the highs or the lows.”
- Dean Witter

"Put all your eggs in one basket and then watch that basket."
- Andrew Carnegie

User avatar
TopComm
Posts: 78
Joined: Thu Jun 03, 2010 7:02 pm

Re: What's the Risk?

Post by TopComm »

Did some tweaking for a little better 7σ with 16198 - Lowest return: 25.99

Post Reply

Fund Prices2024-03-28

FundPriceDayYTD
G $18.15 0.05% 1.05%
F $19.08 -0.06% -0.74%
C $82.21 0.11% 10.55%
S $82.43 0.30% 6.92%
I $42.57 -0.24% 5.95%
L2065 $16.38 0.02% 8.37%
L2060 $16.39 0.02% 8.38%
L2055 $16.39 0.02% 8.38%
L2050 $32.73 0.01% 6.95%
L2045 $14.91 0.02% 6.58%
L2040 $54.38 0.02% 6.22%
L2035 $14.34 0.02% 5.79%
L2030 $47.67 0.02% 5.38%
L2025 $13.15 0.03% 3.43%
Linc $25.61 0.03% 2.82%

Live Charts

Pending Allocations

Under development. For now, you may view Pending Allocations by going to "fantasy TSP" and selecting "Leaderboard sort" of "Pending Allocations".