What Kind of Drawdowns Should We Expect From Saul’s Investing Style?
Yesterday was a hard day. The portfolio dropped roughly 10% in a single session. Watching that kind of red on the screen, even when you think you know what you own, is not easy. It’s the kind of day that makes you ask: Is this normal? Should I be worried? Am I doing something wrong?
Those questions sent me down a rabbit hole — and the result is the analysis I’m sharing here, built on Saul’s complete portfolio record from 2016 to 2024.
I want to be clear about why I wrote this: we need to know the expected drawdowns of this investing style to be mentally prepared. If you don’t know what’s normal, every dip feels like a crisis. If you know that a strategy with a +27% CAGR also carries a -68% drawdown over 34 months, then a bad day, a bad month, even a bad year stops feeling like a personal failure and starts feeling like the cost of admission.
1. The Portfolio at a Glance
First, the raw facts. 100 months of independently verified holdings (Aug 2016 – Dec 2024), cross-checked against Saul’s original Motley Fool posts.[1] 90 unique tickers, 82 with full price history.
| Stat | Value |
|---|---|
| Months tracked | 100 (Aug 2016 – Dec 2024) |
| Unique stocks held | 90 |
| Avg positions per month | 8.6 |
| Longest single hold | DDOG — 43 months (Sep 2019 → Mar 2023) |
| Highest single weight | AXON 35.8% (Dec 2024) |
| Avg monthly turnover | ~14% (Weighted Overlap ~86%) |
The portfolio evolved through distinct eras. Here’s the shape of it:
| Era | Core Positions | Theme |
|---|---|---|
| 2016–2017 | SHOP, LGIH, ANET, PAYC, SQ | Diversified early bets |
| 2018–2019 | AYX, OKTA, TWLO, ZS, MDB, TTD | Pure SaaS basket |
| 2020–2021 | CRWD, ZM, DDOG, NET, UPST, MNDY | COVID SaaS boom |
| 2022 | SNOW, DDOG, BILL, NET, S, CRWD | Bear market survivors |
| 2023 | IOT, MNDY, ELF, CELH, AXON, AEHR | Pivot to new leaders |
| 2024 | NVDA, AXON, NU, S, IOT | Concentration + AI |
This is not a static style. Saul rotated through entire themes as the opportunity set changed — from diversified early bets, to pure SaaS, to post-COVID survivors, to AI and semiconductors.
2. Sector Allocation — The Big Picture
This chart tracks the percentage of the portfolio allocated to each sector, month by month. The dashed line at 100% marks full investment — anything below is cash or margin.
What you see is a story in three acts:
-
2016–2017: The diversified beginning. SaaS, Consumer (SKX), Hardware (ANET), Healthcare, even Real Estate (LGIH). No single sector dominated. Saul was still figuring it out.
-
2018–2021: The Pure SaaS Era. SaaS + Cybersecurity grew from ~50% to 85–95% of the portfolio. Saul concentrated almost entirely into cloud software. This was the bet that made the portfolio — and the bet that would eventually break it.
-
2023–2024: The Great Rotation. Healthcare and Consumer vanished. Semiconductors (NVDA) entered. Fintech grew (NU). SaaS was displaced as the dominant theme. By end of 2024, only 5 names remained.
The gap below 100% in 2023–2024 is Saul systematically withdrawing cash into his “permanent safety fund” — not missing data. This matters: even Saul, the ultimate conviction investor, decided to take chips off the table.
3. The Stocks Themselves — What the Price Charts Reveal
This is the most information-dense chart in the report. Each line is a stock Saul held — solid means he was actually holding it, dashed means 1 year before or 2 years after his holding period. All prices normalized to 100 at first purchase, plotted on a log scale so 100→200→400 are equally spaced. The black line is the S&P 500. The brown shading is the Fed Funds rate.
You can see the COVID trade exploding in 2020 — ZM, CRWD, DDOG, FSLY ran 300–800% from their purchase points, then gave most of it back. You can see the 2022 massacre where nearly every SaaS line collapsed 50–70%. You can see NVDA’s lonely climb in 2023–2024, and ELF and CELH’s brief, spectacular consumer runs.
3.1 Saul’s Entry and Exit Behavior: The Data
I analyzed every “serious” position — a stock first disclosed at >5% of the portfolio. 47 such positions across 2016–2024.
| Metric | Value |
|---|---|
| Total serious positions | 47 |
| Positive return | 29 (62%) |
| Negative return | 18 (38%) |
| Average total return | +61.7% |
| Median total return | +21.7% |
| Average CAGR | +52.2% |
| Median CAGR | +25.7% |
| Sold within 10% of peak | 16/47 (34%) |
| Average % below peak at exit | -21.4% |
| Median % below peak at exit | -19.0% |
That “34% sold near the peak” number looks encouraging — but it’s misleading. Half of those 16 were one-month holds at zero return — stocks Saul tried briefly and dropped immediately. The real near-top exits were mostly moderate winners (COUP +102%, SPLK +60%, TLND +47%, LGIH +90%).
3.2 The Biggest Winners Were Sold Far Below Their Peaks
Every major winner in Saul’s portfolio followed the same pattern: explosive rise → held through the peak → sold deep into the decline. This is not a criticism — it’s a structural property of the strategy.
| Stock | Exit Return | Peak Return | Sold How Far Below Peak? |
|---|---|---|---|
| ZM | +390% | +646% | -34% |
| OKTA | +396% | +582% | -27% |
| AYX | +379% | +620% | -34% |
| SQ | +371% | +482% | -19% |
| SHOP | +258% | +320% | -15% |
| DDOG | +112% | +491% | -64% |
| NET | +13% | +448% | -79% |
| UPST | -13% | +258% | -76% |
| MNDY | -28% | +16% | -38% |
| SNOW | -36% | +44% | -56% |
DDOG is the defining case. Saul held it for 43 months — his longest-ever position. Entry at $33, peak at $197 (+491%), exited at $71. He gave back 64% of the peak gain. Final return was still +112%. UPST and NET are the worst cases: both were massive winners on paper (UPST peaked at +258%, NET at +448%) but ended at -13% and +13% respectively after the full collapse.
3.3 The Real Pattern: “Buy Early, Hold Through Chaos, Sell Late”
The narrative “Saul buys high-growth and rides 5–10x winners” is incomplete. The more accurate description is:
-
Entry timing is excellent. Saul identifies high-growth compounders early — OKTA, AYX, ZM, SHOP, SQ were all bought near their inflection points when they were much smaller and less well-known.
-
He almost never sells at the top. The median exit is -19% below peak. For the biggest winners, the decline is far worse: -27% to -79% below peak.
-
He sells when fundamentals deteriorate, not when price peaks. His posts consistently describe selling because growth rates decelerated, competitive dynamics shifted, or earnings disappointed — events that typically happen after the stock has already declined significantly. He’s a fundamentals-driven seller, not a price-driven one.
-
The late exits appear to be part of the package. It is difficult to see how one captures OKTA at +396% or AYX at +379% without holding through the -30% drawdowns that preceded the eventual exit. If Saul sold at the first sign of a top, he would have exited OKTA at roughly +50% instead of +396%, AYX at roughly +80% instead of +379%. The data strongly suggests that the conviction which generates the upside is the same conviction that holds through the decline — though I want to be careful here: this is an interpretation of the observed pattern, not a mathematical proof. A proper counterfactual analysis (simulating what happens with different exit rules) would be needed to establish causality.[2]
-
When fundamentals never recover, the same conviction becomes costly. NET (+448% peak → +13% exit), UPST (+258% → -13%), and DDOG (+491% → +112%) are cases where the thesis broke and the recovery never came. But even here, the structural approach still produced positive returns in most cases.
The core lesson: Saul’s edge is in selection and conviction, not in timing. He doesn’t try to call tops. He rides the thesis until the fundamentals tell him otherwise. That produces extraordinary winners — and also means he gives back a significant fraction of every peak. The data shows these two outcomes travel together.
4. The Backtest — What $100 Became
This is the simulated portfolio value from $100 starting capital (Aug 2016), rebalanced monthly to Saul’s exact reported weights. Red = Saul, Blue = S&P 500, Green = NASDAQ, Orange = TQQQ (3x leveraged QQQ, included for context — see note below), Purple = Berkshire Hathaway. Log scale. Major drawdowns labeled in red.
A note on the TQQQ comparison: TQQQ is a 3x daily-leveraged ETF. Its returns reflect structural leverage, not stock selection skill. I include it mainly to provide a “maximum leverage” reference point — but for evaluating a stock picker like Saul, the relevant comparisons are QQQ, the S&P 500, and BRK-B.
Returns Summary
| Measuring | Saul | QQQ | S&P 500 | BRK-B |
|---|---|---|---|---|
| Final value (from $100) | $768 | $464 | $270 | $303 |
| Total return | +668% | +364% | +170% | +203% |
The journey, step by step:
-
2016–2018: $100 → $312. Early SaaS bets pay off massively. 3x in 2 years.
-
2019: Flat year. $312 → $357. Saul’s worst pre-COVID year as SaaS multiples compressed.
-
2020 (COVID): $357 → $1,045. The blow-off top. ZM, CRWD, DDOG, FSLY exploded. $100 became $1,045 in 4 years.
-
2021 (Peak): Hit $2,076 in Sep 2021. 20x in 5 years. UPST alone went from 5% to 27% of the portfolio as the stock ran from $40 to $400.
-
2022 (Crash): $2,076 → $1,214. The SaaS bubble burst. -42% in one year. Saul held through the entire decline.
-
2023–2024 (Grind down): $1,214 → $769. Continued bleeding as high-growth stocks re-rated in the rising rate environment. Ended 2024 at 37% below the 2021 peak.
Important caveat on the backtest: The simulation assumes execution at exact month-end closing prices. In reality, Saul posted his holdings at month-end but may have traded intra-month or near the beginning of the following month. For fast-moving names (ZM, FSLY, DDOG in 2020), even a few days’ difference in execution price could move the CAGR by a couple of percentage points in either direction.[3] I have not yet run a sensitivity analysis with T+1, T+3, and T+5 execution assumptions — that is on the to-do list. The 27.1% CAGR should be understood as the best estimate given available data, not a precise number.
Major Drawdowns (Saul)
| Period | Drawdown | Cause |
|---|---|---|
| Nov 2021 – Aug 2024 | -68.5% (34 months) | SaaS bear market + rate hikes. Peak $2,311 → $728. Had not recovered as of Dec 2024. |
| Mar 2020 | -38.2% (1 week) | COVID crash. Recovered in 5 weeks. |
| Sep–Oct 2019 | -37.0% | Growth sell-off. Recovered in 5 months. |
| Feb–Mar 2021 | -28.1% | SaaS rotation. Recovered in 3 months. |
| Oct 2018 | -26.5% | Q4 correction. Recovered in 3 months. |
The 2021–2024 drawdown is the defining challenge of Saul’s strategy: a 34-month, -68.5% drawdown from which the portfolio had not yet recovered as of December 2024. For comparison, the S&P 500’s worst drawdown in the same period was -25% and fully recovered within a year. This is a fundamentally different experience.
5. Performance Metrics — The Full Picture
All computed from monthly returns, Sep 2016 – Dec 2024. TQQQ included in this table for completeness, but the relevant comparisons for a stock picker are QQQ, S&P 500, and BRK-B.[4]
| Metric | Saul | QQQ | TQQQ | BRK-B | S&P 500 | Description |
|---|---|---|---|---|---|---|
| CAGR | 27.1% | 20.3% | 39.7% | 14.1% | 12.7% | Annualized compound return |
| Max DD | -68.5% | -32.6% | -79.1% | -24.3% | -24.8% | Peak-to-trough maximum loss |
| Sharpe | 0.8 | 1.1 | 0.9 | 0.8 | 0.8 | Return per unit of total risk |
| Sortino | 1.4 | 1.6 | 1.3 | 1.5 | 1.2 | Return per unit of downside risk |
| CAPM Alpha (ann) | +11.2% | — | -11.1% | +3.2% | -2.4% | Excess return beyond what market beta predicts[5] |
| Beta | 1.0 | — | 3.1 | 0.6 | 0.8 | Sensitivity to QQQ movements |
| IR vs QQQ | 0.3 | — | 0.8 | -0.3 | -0.9 | Active return consistency vs QQQ |
| Up Capture | 110.9% | — | 288.1% | 62.6% | 72.7% | Performance vs QQQ in up months |
| Down Capture | 58.8% | — | 331.4% | 49.9% | 81.8% | Performance vs QQQ in down months |
Key takeaways from these numbers:
-
Saul beats all equity benchmarks on absolute return. The +27.1% CAGR (2016–2024) and +22.9% CAGR (self-reported 1993–2024) represent extraordinary outperformance over long periods.
-
Saul’s Sharpe is lower than QQQ’s (0.8 vs 1.1). On a risk-adjusted basis, QQQ was the more efficient investment over this specific period. Saul generated higher returns, but with disproportionately higher volatility (11.3% monthly σ vs 5.5%).
-
CAPM alpha of +11.2%/yr is large. This means the portfolio’s returns exceeded what its market beta alone would predict by ~11% per year. However — and this is important — CAPM alpha captures excess return relative to a single-factor (market) model. It does not isolate “stock picking skill” from other potential explanations. Saul’s portfolio has heavy exposure to growth, momentum, small-cap, and SaaS-sector factors. A Fama-French or Carhart multi-factor decomposition would almost certainly attribute some portion of this 11.2% to these factor exposures. I have not yet run those decompositions — it’s high on the list of follow-up work.[5:1]
-
The drawdown problem is equally real. -68.5% vs -32.6% is a fundamentally different risk profile. A 68% drawdown over 34 months means the portfolio must gain +216% from the trough just to break even — a recovery that had not occurred as of December 2024.
5.1 The Longer View: Annual Data 1993–2024
Beyond the monthly data from 2016, Saul has self-reported annual returns going back to 1993 — when he “started seriously keeping track every week.”
| Period | Evidence Quality | Data Source |
|---|---|---|
| 1993–2016 | Self-reported | Saul’s own annual return numbers (unverified) |
| 2016–2024 | Independently reconstructed | 100 months of holdings verified against original Motley Fool posts |
1993–2024 (31 years of returns, vs S&P 500):[6]
| Metric | Saul | S&P 500 |
|---|---|---|
| CAGR | 22.9% | 8.5% |
| Max DD | -68.4% | -40.1% |
| Cumulative | 596x | 13x |
$1 invested with Saul in 1993 became $596 by 2024. The same dollar in the S&P 500 became $13. The gap — 45x — is the cumulative result of 14.4% annual outperformance compounded over three decades.
2000–2024 (25 years, Benchmark = QQQ):
| Metric | Saul | QQQ | S&P 500 | BRK-B |
|---|---|---|---|---|
| CAGR | 20.5% | 7.8% | 5.7% | 10.6% |
| Max DD | -68.4% | -58.3% | -38.5% | -32.1% |
| Cumulative | 105x | 7x | 4x | 12x |
| CAPM Alpha (vs QQQ) | +19.1%/yr [5:2] | — | -0.8% | +6.9% |
| Up Capture | 165% | — | 61% | 59% |
| Down Capture | -14.6% | — | 61% | -3% |
The Down Capture of -14.6% is striking — it means in years when QQQ was down, Saul’s average return was positive. Of the 6 years QQQ declined (2000–2002 dot-com, 2008 financial crisis, 2018, 2022), Saul made money in 4 of them:
| QQQ Down Year | QQQ | Saul |
|---|---|---|
| 2000 | -36.1% | +19.4% |
| 2001 | -33.3% | +46.9% |
| 2002 | -37.4% | +19.7% |
| 2008 | -41.7% | -62.5% |
| 2018 | -0.1% | +71.4% |
| 2022 | -32.6% | -68.4% |
The two exceptions — 2008 and 2022 — were systemic financial crises where correlations converged to 1. The pattern is consistent with stock selection generating returns largely independent of market direction in most environments — though I should note that annual data alone cannot definitively rule out some element of fortuitous timing within each year. The fact that the pattern held across four separate down-years (2000, 2001, 2002, 2018) spanning two different decades does strengthen the case that it reflects something structural rather than luck.
A note on the 1993–2016 data: Saul’s self-reported returns are the only source for this period. They are not independently verified. More importantly, Saul’s fame on this board originated from his success in 2017, 2018, and 2020 — meaning there is a potential survivorship bias in whose track records we analyze. If Saul’s strategy had failed in 2016–2017, we likely would not be studying his portfolio today. This doesn’t invalidate the data, but it’s an important lens through which to view the long-term numbers.[7]
6. Why Saul’s Drawdowns Are So Much Worse Than QQQ’s (Despite Beta = 1.0)
This was the most counterintuitive finding of the entire analysis. Beta = 1.0 implies Saul should move roughly 1:1 with QQQ on average. Yet his max drawdown is 2x deeper and lasted 3x longer. The answer lies in what Beta does NOT capture.
6.1 Beta Is Insufficient, Not Wrong
CAPM Beta only measures average co-movement with the market over the entire sample. It says nothing about volatility magnitude (Saul’s monthly σ = 11.3% vs QQQ’s 5.5%), concentration risk (5–10 stocks vs 100), regime dependence, or how long drawdowns last. The risk is regime-dependent, and a single number can’t capture that.
6.2 Regime-Dependent Beta
We decomposed beta by market condition. Important caveat: the sub-samples here are small (n=14 and n=15 for the extreme/mild splits). The point estimates below should be understood as suggestive patterns, not precise measurements — confidence intervals would be wide.[8]
| Market Regime | Beta | R² | n | Meaning |
|---|---|---|---|---|
| QQQ up months | 1.57 | 0.22 | 67 | Saul amplifies gains by 57% |
| QQQ down months (all) | 0.66 | 0.05 | 33 | Almost no linear relationship — idiosyncratic |
| QQQ mild down (0 to -3%) | 6.03 | 0.23 | 15 | Extremely sensitive to small dips (wide CI) |
| QQQ extreme down (<-5%) | 3.35 | 0.42 | 14 | Crashes harder than QQQ (wide CI) |
The key insight: in down months, R² = 0.05. Saul’s returns during QQQ down months are almost entirely idiosyncratic — they’re driven by stock-specific events, not by the magnitude of QQQ’s decline. In mild down months, high-beta SaaS names actually fall harder than QQQ (beta = 6.03). In extreme crashes, everything falls together (beta = 3.35). But the low R² means these relationships are noisy — stock-specific disasters can happen in any market environment.
6.3 Down Capture Decomposition: The Hidden Pattern
Overall Down Capture = 58.8% suggests Saul loses 41% less than QQQ in down months. But this average masks a crucial asymmetry:
| Down Month Type | n | QQQ Avg | Saul Avg | Down Capture | Who Does Better? |
|---|---|---|---|---|---|
| Mild (0 to -5%) | 19 | -1.75% | -2.81% | 161% | QQQ does better |
| Extreme (<-5%) | 14 | -7.95% | -2.25% | 28% | Saul does better |
| All down months | 33 | -4.38% | -2.57% | 58.8% | Saul does better |
As with the regime beta, the extreme-month sub-sample is small (n=14). One or two outlier months (e.g., March 2020) can meaningfully shift the “extreme” average. This is a genuine pattern worth monitoring, but it should not be treated as a stable parameter.
The overall 58.8% is dominated by the extreme months where QQQ’s losses are much larger in magnitude. But in mild down months, Saul actually loses MORE than QQQ (161% capture — loses $1.61 for every $1.00 QQQ loses). This is the mechanism behind the paradox: Saul does not lose more money in market crashes. He loses money in the OTHER months — when QQQ is flat, mildly down, or even up — because his concentrated positions have their own blowups.
Proof: 9 months where Saul lost >10% while QQQ was above -3%:
| Date | Saul | QQQ | Gap |
|---|---|---|---|
| 2019-09-30 | -23.0% | +0.9% | -23.9% |
| 2021-11-30 | -15.2% | +2.0% | -17.2% |
| 2021-12-31 | -13.1% | +1.2% | -14.2% |
| 2022-05-31 | -20.1% | -1.6% | -18.5% |
| 2022-11-30 | -16.3% | +5.5% | -21.9% |
| 2023-04-30 | -12.9% | +0.5% | -13.4% |
| 2023-10-31 | -21.0% | -2.1% | -18.9% |
| 2024-06-30 | -11.4% | +6.5% | -17.9% |
| 2024-07-31 | -11.5% | -1.7% | -9.8% |
These are NOT market-driven losses. They are stock-specific blowups — earnings misses, guidance cuts, sector rotations. Beta cannot capture this risk. CAPM’s R² for Saul’s entire return series is only 0.23 — meaning 77% of Saul’s monthly return variance is unexplained by QQQ. However, I should note: low R² does not necessarily imply stock-picking skill. It could also reflect sector factor exposure, style factor exposure (growth/momentum), or simply the noise introduced by high concentration. A proper factor decomposition would help separate these.[5:3]
6.4 The Duration Problem
| Saul | QQQ | |
|---|---|---|
| Worst drawdown | -68.5% | -32.6% |
| Duration | 34 months | 12 months |
| Recovery | Still ongoing as of 2024 | Fully recovered |
The drawdown duration, not just the depth, is what makes Saul’s strategy extreme. A -68.5% decline over 34 months means nearly 3 consecutive years of watching your portfolio shrink, multiple false recoveries that reverse, and QQQ making new highs while Saul still bleeds. This is fundamentally different from QQQ’s acute crash (2022) followed by a clean recovery (2023). Most investors can survive a sharp crash. Very few can survive 3 years of chronic decline.
6.5 The Resolution
Saul’s risk summarized:
| Risk Component | Saul vs QQQ |
|---|---|
| Systematic (Beta) | Similar (β≈1.0) |
| Idiosyncratic volatility | 2x QQQ (σ 11.3% vs 5.5%) |
| In normal down months | Worse than QQQ (Down Capture 161%) |
| In extreme crashes | Better than QQQ (Down Capture 28%) |
| Drawdown depth | 2x QQQ (-68.5% vs -32.6%) |
| Drawdown duration | 3x QQQ (34 vs 12 months) |
The story is not that “Beta is wrong” — it’s that Beta only captures one dimension of risk. Saul’s concentrated, idiosyncratic strategy generates large excess returns relative to its CAPM benchmark, but exposes investors to risks that a single-factor model was never designed to measure: single-stock blowups in benign markets, serial correlation of negative returns, and multi-year drawdowns that test the limits of conviction investing.
7. The Trade-Off
Let’s lay it out cleanly.
| Saul | QQQ | |
|---|---|---|
| Annual return (CAGR) | +27.1% | +20.3% |
| Cumulative return | +668% | +364% |
| Worst drawdown | -68.5% | -32.6% |
| Drawdown duration | 34 months | 12 months |
| Sharpe ratio | 0.8 | 1.1 |
| Monthly volatility | 11.3% | 5.5% |
| CAPM Alpha vs QQQ (annualized) | +11.2% | — |
| Diversification | 5–10 stocks | 100 stocks |
| Variance unexplained by QQQ | Dominant (R²=0.23) | Near zero |
Saul’s extraordinary returns come from a coherent strategy where each element reinforces the others:
-
Stock selection has produced large excess returns. The CAPM alpha of +11.2%/yr (2016–2024) and +19.1%/yr (2000–2024, using self-reported annual data) show returns substantially above what market exposure alone would predict. He identified high-growth compounders near their inflection points — OKTA, AYX, ZM, SQ, SHOP, DDOG were all bought when they were much smaller and less well-known. The data is consistent with genuine selection ability, though as noted above, I have not yet isolated this from factor exposures (growth, momentum, size, sector).[5:4]
-
Conviction and drawdowns appear to be two sides of the same coin. The same holding period that allowed capturing multi-year compounders (OKTA +396%, AYX +379%, SHOP +258%) also meant holding through their eventual declines. The median exit at -19% below peak, and the deeper declines for the biggest winners (-27% to -79%), are not obviously a separate “weakness” — the data strongly suggests they are connected to the holding period that generated the returns. If Saul had sold at the first sign of a top, he would have exited OKTA at roughly +50% instead of +396%. That said, whether a systematic exit rule could improve risk-adjusted returns without sacrificing too much upside is an open empirical question that deserves its own analysis.[2:1]
-
The annual data suggests stock selection works over 12-month horizons. Saul made positive returns in 4 of the 6 years when QQQ declined, including the dot-com bust (2000–2002). This pattern — positive returns in bear markets driven by idiosyncratic selection rather than market direction — is what the monthly data would lead us to expect. That said, annual data cannot definitively rule out intra-year timing effects; the consistency across multiple bear markets separated by decades is suggestive but not dispositive.
The net result: a 31-year track record with +22.9% CAGR (self-reported), $1 → $596, driven by concentrated stock selection and the willingness to hold through extreme cycles. The path volatility (-68.4% max drawdown, 11.3% monthly σ) is the structural consequence of concentration — and concentration is part of what makes the returns possible. Whether there exists a version of this strategy that preserves most of the return while materially reducing the drawdowns is, in my view, the most important open question.
8. How This Was Built — An AI Disclosure
I’ll be transparent here, in the spirit of the disclosures that others on this board have established.
This analysis was produced with the assistance of an Agentic AI system. I designed and built the research agent myself — it scrapes the data, runs the backtests, generates the charts, runs the regressions, and iterates on the analysis. Every number in this report is verified against Saul’s original posts and public market data. But the “analyst” writing the Python, computing the regressions, and drafting the narrative is an AI working under my direction.
Why build this? Because my largest constraint is time. I have a very demanding job — working past midnight is not rare — and there simply aren’t enough hours in the day to do this kind of research manually. But here’s the thing: efficiency increases 10x when you can build your own Agentic AI. What previously would have taken a team to research, backtest, chart, and write up can now be done by one person with the right tools. That’s the only reason this analysis exists, and I suspect it’s going to change how a lot of us do research going forward.
Disclaimer: This post and the accompanying report were generated with AI assistance. All data sources are cited — 100 months of Saul’s verified monthly holdings, stock prices from yfinance and Investing.com, benchmarks including QQQ, S&P 500, and BRK-B. All analysis has been reviewed by a human. But you should apply your own judgment to any AI-assisted content, just as you would to any investment analysis.
9. Thank You
First, to Saul: thank you for the extraordinary generosity of sharing your holdings every month for over a decade. The knowledge embedded in that record is remarkable — a multi-decade track record, 596x cumulative return, a masterclass in conviction investing. I’ve learned more from studying your actual portfolio behavior than from any book or course. What you’ve built — both the returns and this community — is something special.
And to this board: the quality of discussion here, the willingness to share ideas and challenge each other, is rare on the internet. Thank you for keeping it that way.
10. More Questions Than Answers
This analysis opens more questions than it answers. A few that I’ve been thinking about, and hope the board might weigh in on:
-
The -68.5% drawdown over 34 months is the defining challenge. Can we reduce this without sacrificing the alpha? Position sizing rules? Sector concentration caps? Some form of portfolio-level risk management that doesn’t require market timing?
-
The 9 months where Saul lost >10% while QQQ was flat or positive — these are stock-specific blowups, not market events. Is there a way to detect deteriorating fundamentals earlier and exit before the worst of the decline, without becoming a momentum trader?
-
How much of Saul’s alpha comes from the names versus the concentration? If you held the same 20 names at equal weight instead of 8 concentrated, what happens to the Sharpe ratio? Is the concentration essential or incidental?
-
The Down Capture decomposition shows Saul actually loses more than QQQ in mild down months (161%), but far less in crashes (28%). Is there a way to hedge against the “mild down” regime specifically — the grinding stock-specific bleed — while staying fully exposed to the upside and the crash resilience?
-
The annual data shows Saul made money in 4 of 6 QQQ down years. What characteristics did those four successful years share (2000, 2001, 2002, 2018), and what broke in 2008 and 2022? Is there a pattern we can learn from?
Some of these questions are fair game for this board — they’re about stock selection and portfolio construction, which is what we discuss here. Others start to drift into portfolio management or macroeconomics territory, which I recognize is Out of Topic. I’ll be thoughtful about which ones I bring here, and why they’re suitable. When I have time to research them — and now that I have the infrastructure in place, that might be sooner than later — I’ll share what I find.
Data & Methodology
-
Holdings: 100 months verified against original Motley Fool posts
-
Stock prices: 82 stocks, daily adjusted close (yfinance + Investing.com)
-
Benchmarks: QQQ, TQQQ, S&P 500 (^GSPC), NASDAQ (^IXIC), BRK-B, 13-Week T-Bill (^IRX)
-
Caveats: 9 of 90 tickers have no price data (mostly acquired/delisted micro-caps, treated as constant); monthly rebalancing assumes execution at exact month-end closing prices (see footnote 3 for sensitivity discussion); simulation assumes perfect liquidity — real portfolio sizes might have moved markets in small-cap names; backtest does not include dividends, taxes, or transaction costs.
All 2016–2024 holdings have been independently cross-referenced against Saul’s original Motley Fool posts. This is the “verified” portion of the dataset. ↩︎
Counterfactual analysis not yet performed. The claim that late exits are “inseparable” from the returns is an interpretation based on the observed holding pattern. To properly test this, one would need to simulate alternative exit rules (e.g., sell at -20% from peak, sell at -30%, trailing stop) and measure the impact on both CAGR and drawdowns. This is planned follow-up work. ↩︎ ↩︎
Backtest execution sensitivity not yet analyzed. Saul posts holdings at month-end but may execute trades at different times within the month. For volatile names, a few days’ difference in execution price could meaningfully shift the CAGR. I plan to run a sensitivity test with T+1, T+3, and T+5 execution assumptions. The 27.1% CAGR should be interpreted as an estimate with some range around it, not a precise measurement. ↩︎
TQQQ is a 3x daily-leveraged ETF. Its returns reflect structural leverage, not stock selection skill. It is included in the metrics table for completeness, but the relevant comparisons for evaluating a stock picker are QQQ, S&P 500, and BRK-B. ↩︎
CAPM Alpha ≠ pure stock-picking skill. CAPM alpha measures excess return beyond what a single-factor (market beta) model predicts. Saul’s portfolio is heavily exposed to growth, momentum, size, and SaaS-sector factors. A multi-factor decomposition (Fama-French 3-factor, Carhart 4-factor, or FF5) would likely attribute a portion of the 11.2% alpha to these factor exposures. I have not yet run these decompositions. Until then, the alpha should be understood as “excess return relative to CAPM” rather than “pure skill.” I also note that the 19.1% alpha figure for 2000–2024 uses annual data and self-reported returns for 2000–2016; the methodology (regression intercept from annual returns, geometrically annualized) differs from the monthly computation. A more detailed reconciliation of the two alpha figures deserves its own footnote — for now, readers should treat the annual alpha with appropriate caution. ↩︎ ↩︎ ↩︎ ↩︎ ↩︎
The 1993–2024 period spans 32 calendar years but 31 years of returns (1993 through 2024 inclusive). All CAGR calculations use 31 years. ↩︎
Survivorship bias caveat. Saul’s fame and our ability to study his portfolio both stem from his success. If his strategy had failed in 2016–2017, we would likely not be analyzing it today. This is not a critique of the data — we have 100 verified months, which is rare and valuable. But it means the “32-year track record” framing gives equal visual weight to verified data (2016–2024, 8 years) and self-reported data (1993–2016, 23 years). ↩︎
Small sample warning for regime analysis. The mild-down (n=15) and extreme-down (n=14) sub-samples are small. Beta estimates from these samples have wide confidence intervals. The point estimates (β=6.03, β=3.35) should be treated as suggestive of a pattern, not as precisely estimated parameters. A single outlier month can meaningfully shift these numbers. I have not yet computed formal confidence intervals or t-statistics for the regime splits — this is on the follow-up list. ↩︎
