Investment Strategies / Mechanical Investing
No. of Recommendations: 6
Do folks find that subscribing to the 20 year backtest level is useful?
I just ended the free trial which came with 10 year backtest. Tried recreating a few of my GTR1 screens in P123.
It was difficult to compare to GTR1 since current data is not available in GTR1.
Seems to me that more backtest data is always preferable, on the other hand its relatively expensive.
And GTR1 can be used for backtesting, if translating screens between the two was known to "work".
Mark
No. of Recommendations: 5
Do you need 20 years at Portfolio123? Probably not but I’ve found that the ranking systems based on at least 10 years of backtesting have a higher probability of post discovery performance. More importantly the minimum P123 level that gives you the real advantages of P123 is “Backtest” at $1000 / year which gives you 10 years of backtest and most important Custom Ranking Systems. Ranking systems are much superior to Screeners! Their Ultimate Level for a non-professional individual user is $2400/yr and AI tools cost another $1200 year plus additional computer charges. Another 40% for Europe and UK stocks. Professional users pay significantly more.
The P123 level that is best for an individual depends on three things, first the amount of money they are investing and second the amount of time they are willing to invest and last the percentage increase in return they believe they can achieve.
No. of Recommendations: 6
TGMark wrote:
"It was difficult to compare to GTR1 since current data is not available in GTR1." I recently signed up for the P123 trial plan, and am running through my evaluation steps. I expect P123 will pass all of my tests, but could post details if people are interested. Let me know.
For comparison between GTR1 and P123, I plan on looking at the correlation of monthly excess returns (monthly return minus benchmark return). I will also compare CAGR across multiple screens. I expect P123 and GTR1 to match close enough for my purposes.
TGMark wrote:
Do folks find that subscribing to the 20 year backtest level is useful? I need at least a 10-year backtest for my WWL system, and 20 years for new screens. But, maybe AI will change that.
How Far Back Should You Backtest?, February 15, 2025
"The result confirmed what I’d been testing for the last eight years or so: if you’re choosing between a backtest period of three to fifteen years, a nine-to-twelve-year lookback is optimal. ... The lookback periods after fifteen years begin to show increasing correlation, until you reach 24 years, at which point the correlations exceed those of the nine-to-twelve-year lookback."
https://blog.portfolio123.com/how-far-back-should-...
No. of Recommendations: 5
Thx all for the responses. Right now P123 is telling me there's a 25% discount for the first year for the 20-year backtest plan.
Instead of $2400/yr it's $1800. Not so bad a premium over the 10 year plan.
For comparison between GTR1 and P123, I plan on looking at the correlation of monthly excess returns (monthly return minus benchmark return). I will also compare CAGR across multiple screens. I expect P123 and GTR1 to match close enough for my purposes.
I am interested to see what you find. For screens that are picking stocks, I tend to want to see pretty good agreement in which tickers are picked.
That is a tall order, given the differences in universes (non-OTC at P123 vs GTR1), lags, and myriad other details.
I did not get what I consider to be great agreement between P123 and GTR1 for the screen I worked on the most.
It had decent CAGR in both, but the picks were almost completely different.
And I read that blog post on backtest length, and saw that P123 usually considers 10 years to be near optimum.
Seemed odd - here, it's always been "the longer the better" which makes intuitive sense.
Mark
No. of Recommendations: 4
Lizgdal: I need at least a 10-year backtest for my WWL system, and 20 years for new screens. But, maybe AI will change that.
I used a WWL system for several years, my logic was what’s working lately that has also worked long term which needs both recent and long term to make the decision. I now have a whole series of AI simulations using 20 years of data where I train multiple models for 4 to 17 years, after a gap to prevent data leakage verify performance for the following year. Outperformance in general but not guaranteed begins after 5 or 6 year of training.
Your link to Yuval Taylor’s P123 blog above is a good example of how much effort he puts into his studies.
His end of year 2025 post on his performance.
“””I had an excellent year.
My hedge fund, Fieldsong Investments, was up 59.54% for the year. My private foundation was up 24.94%. Both of these use leverage and are hedged.
My wife's and my retirement accounts that are not invested in the hedge fund were up 34.59%. My kids' accounts were up 59.60%. These accounts don't use leverage and were not hedged.
Altogether I made $3.93M. And all of this was using Portfolio123 as the basis for all of my decision-making. Thank you P123!
I also manage two other accounts using Portfolio123 and IB. One was up 60.91% and the other was up 28.21%.
“””
How did your 2025 P123 portfolio perform? - General - Portfolio123 Community
I’ve been using P123 for years and believe it is the best site available for factor investing however only a small percentage of the investors there do as well as Yuval. But Yuval is obviously a very intelligent individual and has been doing this full time for several years. The more successful individuals tend to have either a strong background in statistics, finance or software.
No. of Recommendations: 12
P123 usually considers 10 years to be near optimum.
Seemed odd - here, it's always been "the longer the better" which makes intuitive sense.
10 years back is 2016. That period has been mostly a bull market without ANY bear market. Just a couple of shortish corrections.
Any backtest which doesn't include at least the 2008/2009 bear market will be telling you sweet nothings.
No. of Recommendations: 10
I’ve been using P123 for years and believe it is the best site available for factor investing however only a small percentage of the investors there do as well as Yuval. But Yuval is obviously a very intelligent individual and has been doing this full time for several years. The more successful individuals tend to have either a strong background in statistics, finance or software.
This is a prime example of survivorship bias.
You never hear of the very intelligent individuals who used P123 and got only so-so returns.
Nobody brags "I dropped $2400 on P123 and managed to match the S&P 500."
No. of Recommendations: 3
10 years back is 2016. That period has been mostly a bull market without ANY bear market. Just a couple of shortish corrections.
Any backtest which doesn't include at least the 2008/2009 bear market will be telling you sweet nothings
Paraphrasing, all good stocks go up for the same reasons, poor performing stocks go down for different reasons.
Would a backtest that included the Dotcom bust be relevant to the GFC bust? I have not checked to see if this is true. Does anyone have insight on this. Having now just contemplated RayVt's statement, I am not sure including the GFC decline helps selection of a screen that will perform in the next major decline. I can only think that something like BCC protects you in a decline and then select screens that do well in a bull market.
Aussi who has used backtests over the longest possible period
No. of Recommendations: 7
Does anyone have insight on this
Not really. One crisis does not seem to be greatly predictive of the next one.
The tech bubble was different than the GFC which was different than the pandemic.
Nothing really "worked" through the depth of those bears.
You had to be diversified enough to weather them adequately, or just grit it out and keep going.
Or follow some timing that gets you out before the bottom and gets you back in when there's blood in the streets.
Mark
No. of Recommendations: 4
Rayvt: “Any backtest which doesn't include at least the 2008/2009 bear market will be telling you sweet nothings.”
I wouldn’t say sweet nothings, it wouldn’t protect you from another 2008 but it would have still helped you outperform at later dates. I just looked at how my ML/AI models trained from 2004 till end of 2007 did in 2008. You are right they didn’t outperform but neither did they underperform. Likewise when the covid crash happened my ML models went down just as fast as the rest of the market but the interesting observation it one of the best performance periods happened in the following recovery. It appears the ML models found mispriced equities when they were all down.
Ravet: “This is a prime example of survivorship bias. You never hear of the very intelligent individuals who used P123 and got only so-so returns.”
Good point. P123 has what they call designer models developed by individuals who believe they have found an outstanding system which you can subscribe to. The vast majority of these did not perform better than the SP500 post discovery, especially in the early period. So, I believe your observation appears to be correct. Developing a ranking system that outperforms is not an easy couple of hours a week project.
My point is P123 along with their community have developed tools to analyze how hundreds of individual factors have performed over the last 20 years, measure how much information they give you and how they correlate to other factors. Professional investment managers don’t pay out $6K a year for multiple years if they weren’t satisfied. Perhaps that isn’t the best logic as many more pay > $30K a year for a Bloomberg terminal.
No. of Recommendations: 4
I am not sure including the GFC decline helps selection of a screen that will perform in the next major decline. I can only think that something like BCC protects you in a decline and then select screens that do well in a bull market.
Absolutely.
These are wo different things.
I just find it annoying and not helpful when somebody presents
"WOO-HOO!!! This screen does fantastic (in a period which is a bull market)."
Duh. Just about everything does great in a bull market.
If a backtest DOES include at least one bear market and a couple of correction markets, *then* it attracts my interest.
But most people don't use any sort of robust timing scheme to sidestep the full depth of a bear market.
That's assuming that there IS such a timing scheme.
And I don't generally come across a backtest that even contemplates a timing overlay.
Usually when I want to do that, I download (if possible) the daily/weekly/monthly values of the backtest into a spreadsheet and then apply my own timing scheme to it.
No. of Recommendations: 1
And I don't generally come across a backtest that even contemplates a timing overlay.
Is timing a thing on P123 like it is here?
No. of Recommendations: 11
lizgdal wrote: "For comparison between GTR1 and P123, I plan on looking at the correlation of monthly excess returns (monthly return minus benchmark return). I will also compare CAGR across multiple screens. I expect P123 and GTR1 to match close enough for my purposes."
TGMark wrote: "I am interested to see what you find. For screens that are picking stocks, I tend to want to see pretty good agreement in which tickers are picked." I've finished my testing, and will post details in a separate thread (maybe in about a week). P123 and GTR1 match closely enough for my purposes. I mostly avoid nano caps (market cap < $50 million), and so did not do much testing on those. I found some implausible returns, but those are mostly avoided with the usual liquidity criteria (e.g. Price above 1).
The tickers are unlikely to match. I would expect about 50% agreement on average, with a range from 0% to 100% agreement. Even something as simple as Market Cap is difficult to match.
=== links ===
Important Update: Transition to FactSet Market Cap, Feb 6, 2026
https://community.portfolio123.com/t/important-upd...Author: borisnand
Subject: Comparing GTR1 picks and SI Rankings
Date: 6/1/2016
"Overall, there is good agreement between the Rankings and GTR1, with 56% Jaccard index on average."
Jaccard index = (number of picks common to both lists) / (number of picks in combined list)
or put another way:
Jaccard index = size of the intersection divided by the size of the union of the sample sets
https://yorickm.com/Message.php?pid=32264724Author: borisnand
Subject: Comparing GTR1 results and Weekly Trade Ledgers
Date: 6/1/2016
screen CAGR_L CAGR_G GSD Jaccard
Fried_500 17% 14 31 100%
OptiMan 7% 5 28 100%
Steady_Growth 3% 1 21 100%
Bob 15% 13 25 96%
P/SLove_You 11% 12 27 92%
PIH_Naked 11% 12 32 92%
CANSLIM-26 5% 4 21 90%
Gentle_Screamers 21% 22 26 82%
GS_Mungo 15% 9 22 82%
GS_Mungo_Voom 14% 8 22 82%
Microcap_Momentum 31% 17 43 79%
POG 8% 7 37 79%
Shrinkage 9% 11 26 79%
Turnarounds 4% 6 36 79%
Small_Value 12% 8 32 76%
FCF-26 13% 6 30 72%
Quality_Earnings 12% 10 32 72%
Melange 5% 10 36 67%
Net-Nets_Grahamified 22% 21 30 64%
POI 19% 18 37 61%
Silver_Parachute 11% 13 26 61%
Value_at_the_Top 12% 12 19 61%
ARS 7% 8 24 56%
GSX 13% 15 23 47%
GSX2 13% 19 24 43%
Zweig-26 1% -4 20 43%
S&P_Peg 3% 4 29 19%
Z26saTA -1% -2 22 18%
Up5X3 29% 23 32 15%
Low_Mult 21% 27 24 9%
78RPM 12% 14 34 4%
High_Relative_Value 25% 30 21 4%
Up_5% 12% 17 33 4%
WK_Voom 17% 20 26 0%
average 13% 12 28 60%
https://yorickm.com/Message.php?pid=32264753Author: rrjjgg
Subject: Re: GTR1 verification needed
Date: 3/19/2017
"I`ll just assume that [Microcap_Momentum] is too chaotic to be easily backtested and should be avoided."
https://yorickm.com/Message.php?pid=32642718
No. of Recommendations: 8
I am mostly interested in monthly screens, long only, no hedging or timing, and no nano caps. So I did not do much testing on:
penny stocks (less than $1 Price)
nano caps (less than $50 million Market Cap)
daily or weekly trading
timing or hedging
ADR, MLP, ETF, or CEF
non-US stocks
My Base screen is:
GTR1 {gBase}
step0: Ordinary common stock or REIT.
step1: [SI Sales Q1] > 0
step2: [Mkt Days Since Security Opened] >= 253
step3: [Actual closing Price] > 1
step4: [Average dollar-volume over 63 days] > 1,000,000
step5: [MktCap] >= 50
step6: Last Q report filed within the past year.
step7: Only one ticker per company.
Rebalance every 19 mkt days, equal weight.
P123 {pBase}
Universe(PRIMARYUSA)
!IsMLP & !IsOTC & !IsUnlisted
SalesQ > 0
Close(253) > 0
Close(0) > 1
AvgDailyTot(63) > 1000 * 1000
MktCap >= 50
BarsSince(PeriodDateQ) <= 252
!StaleStmt
Rebalance every 4 weeks, no hedging, equal weight.
PRIMARYUSA is US equities excluding ADRs and foreign stocks.
