A backtest number only means something once you know what it hides.
Total return looks impressive on its own. Maximum drawdown decides whether you could have held the position at all. Here we explain what each figure means, how they mislead, and how to design a split-buying strategy that survives a real decline.
Open the backtesting tool →
Backtest by ticker
Each ticker has its own page with ten years of split-buying results, a year-by-year table and the things worth knowing before you test it.
Start here
What Is Split Buying? How to Use This Backtest Tool
Split your capital, buy the dips, sell each tranche at its own target. Every input explained, plus how to read the results.
What Is Maximum Drawdown? Why a 50% Loss Needs a 100% Gain
Recovering from a 50% loss takes a 100% gain. What drawdown is, and why you should read it before the return figure.
Infinite Buying V4.0 backtest — 16 years of TQQQ and SOXL, plus the dot-com and 2008 crashes
We rebuilt Laoer's Infinite Buying method V4.0 on daily bars and ran it on TQQQ and SOXL for 2011–2026 and a synthetic 1999–2010, against holding and DCA.
More guides
Backtesting SOXL and TQQQ back to the 1990s — index-based history built from the 1x index
SOXL and TQQQ listed in 2010. To test them through the dot-com bust and 2008, we built pre-listing prices from the 1x index. How it is done, and how closely it matches the real ETFs.
We Found the 'Perfect' Partner to Cover SOXL's Bad Years. Then 30 Years of Data Broke All Six
Adding SOXS, FAS or NAIL strategies next to a SOXL ladder visibly cut drawdowns in 2016–2026. Run the same rules on the decades before they were built, and every one of the six lost its edge. Here is what the illusion was, in charts.
If SOXL Had Existed in 1996 — Would Ladder and Dual-Mode Strategies Have Survived the Dot-Com Crash and 2008?
SOXL listed in 2010 and has only seen the best 16 years in chip history. We rebuilt it from semiconductor data back to 1996 and ran the dual-mode (D1–D5) strategies and today's site strategies. Run continuously through the dot-com bust and the financial crisis, drawdowns reached −40% to −93%. Recalculated on 2026-10-19 with the new index-based data.
Total Return vs CAGR: Why +50% Then −50% Leaves You at −25%
Up 50% then down 50% leaves you at −25%, not flat. The compounding illusion and how to read annualised return.
Do rising rates really hurt semiconductors? We measured it
We split 33 years into hiking, holding and cutting regimes and measured returns by asset. The result contradicts the conventional story.
Do quant strategies only work on charts that go up? 34,000 backtests on leveraged gold ETFs
A split-buy rule earned 23% a year on GDXU while the ETN fell 94% from its peak, yet it badly lagged buy-and-hold on a steadily rising 2x gold ETF. What a strategy needs is not an uptrend.
Can the 10-year Treasury yield predict SOXL? We measured the lead and lag
Correlations between yield changes and 3x semiconductor returns, shifted from -20 to +20 sessions, with R² for every result.
Buy Korea at the open after KORU jumps? A 13-year backtest, and what happens in reverse
Buying KODEX Leverage at the Korean open after a strong night for KORU lost money in all 64 parameter combinations. Where the gain goes, and what the reverse trade looks like.
What to check first when a backtest looks good
A 72% CAGR strategy that fell apart under walk-forward testing. We use our own strategy as the example, and show every step.
Five Reasons a Backtest Looks Better Than Reality
Overfitting, look-ahead bias, survivorship, execution costs and regime dependence. What to verify before you trust a result.
Seventeen leveraged ETFs run through the same rule
Every ticker on this site, same strategy, same seed, same fee — return, drawdown and losing years side by side.
ETN vs ETF: What Happens If the Issuer Walks Away
They trade the same way on your screen, but an ETN is an unsecured promise. If the bank behind it walks away, so does your money. The FNGU case shows how.
What happened after the VIX curve inverted
We gathered the 248 sessions where VIX divided by VIX3M went above 1 and measured the next one and three months.
What tranche weights actually change
The conservative, standard and aggressive presets differ only in weights, target and time cut. Sixteen years of SOXL shows what that difference buys.
Volatility Drag Explained: Why a 3x ETF Doesn't Return 3x
Why a 3x fund does not hand you 3x over time. The arithmetic, a rough formula for sizing the drag, and when leverage works for you instead of against you.
8 Checks Before You Trade a Backtest
The result looks good. Here is what to verify before real money goes in — parameter wobble, costs at 3x, bear markets included, and a month of paper trading.
Removing the time cut multiplies returns fivefold. So why keep it?
We varied the time cut from 3 to 250 sessions. Here is exactly where return and drawdown part company.
Splitting one account into three reduced the drawdown
Why running several uncorrelated strategies side by side lowers drawdown, explained with the actual correlation matrix and results.
How to backtest SOXL — from its 2010 listing to today
A step-by-step procedure for testing the 3x semiconductor ETF: which periods to include, which numbers to read first, and what to verify before trusting the result.
How many times leveraged were 3x ETFs, really? Sixteen years measured
SOXL, TQQQ, UPRO and UDOW against their underlying indices. Daily it was exactly 3x. Long term it was nothing like it.
TQQQ or SOXL — sixteen years of data compared
Nasdaq 3x against semiconductor 3x, run on identical settings. Returns, drawdown, volatility drag and split-buying results side by side.
Does averaging down actually work? Seven tickers, sixteen years
We ran add-on-the-way-down across SOXL, TQQQ and five others for sixteen years. The percentage improves while the money lost grows.
How many tranches should you split into — three to ten
What changes when you add tranches: buying opportunities, drawdown, idle cash, and the whole-share trap when your seed is small.
How Much Should Each Top-Up Be? Three Amounts, Tested
We ran 100%, 50% and 25% of the seed as the top-up amount across 7 tickers and 16 years. You keep 68% of the benefit and 14% of the downside.
Three ways to measure return when you added money midway
Dividing by total capital penalises money added late. Simple return, Modified Dietz and time-weighted return, worked through one example.
Good parameters sit on a plateau, not on a peak
If only one setting looks good, it is fitted to the past. How to tell a plateau from a peak, with real numbers.
What this site covers
Reading the numbers
Maximum drawdown, CAGR, Calmar, Sharpe and win rate. What each measures, and the situations where each one lies to you.
Split-buying strategies
Dividing capital into tranches, buying declines, selling each tranche at its own target, and cutting stale positions.
Where backtests break
Overfitting, look-ahead bias, survivorship, execution costs and regime dependence.
Leveraged products
Why a 3x fund does not deliver 3x over time, and how volatility drag accumulates in sideways markets.