Backtest Disclosure & Performance Reporting

TL;DR. Every backtest number that appears anywhere on this site is hypothetical, computed against historical data, and does not represent the result of an actual investment. This page is the canonical disclosure for our backtest methodology, the assumptions baked in, the limitations we know about, and the regime-dependence that makes any past result a fragile guide to the future. If a number on this site doesn't link to this page, it should — file a bug report.

What our backtests actually do

When you run a strategy backtest on Tessera Alpha, the engine:

1. Reconstructs the eligible universe of US-listed stocks at each weekly rebalance date over the chosen window (default: 10 years), including stocks that have since been delisted, acquired, or renamed (no survivorship bias).
2. Applies the strategy's filter rules using only data that was publicly available as of that rebalance date (no look-ahead bias).
3. Selects positions per the strategy's ranking rule and applies the configured position sizing.
4. Holds positions until the next rebalance, with optional stop-losses, trailing stops, or regime-based exits if the strategy specifies them.
5. Computes returns net of an 8 basis-point round-trip transaction cost assumption (you can override this in the strategy config).
6. Reports CAGR, Sharpe ratio, Sortino ratio, max drawdown, and benchmark-relative statistics versus SPY (and optionally QQQ).

The data underneath is point-in-time accurate — fundamentals are stamped to the date they were reported, not retroactively restated. Earnings restatements that happen after the fact don't leak back into the screen.

What our backtests do not model

• Slippage at execution. We assume you transact at the closing price on the rebalance date. In reality, large orders move prices, especially in small-caps. The 8 bps transaction-cost assumption is intentionally conservative for a retail-sized portfolio (under ~$500K AUM); if you scale up, slippage will degrade returns.
• Bid-ask spreads at small market caps. Stocks below $1B market cap can have spreads of 0.5%–2%. The transaction cost assumption does not separately model this.
• Borrow costs for short positions. Tessera's strategies are long-only by default, so this is not relevant unless you've explicitly built a short leg.
• Tax drag. Backtest returns are pre-tax. Realized gains in a taxable account meaningfully reduce after-tax CAGR; this is not modeled.
• Drift between rebalance dates. Position-level rebalancing happens weekly. Within a week, the engine assumes the portfolio holds proportional to entry weights — no intra-week rebalancing or risk-management overlay.
• Behavioral drag. The single largest gap between backtest CAGR and live CAGR for retail investors is behavioral — sitting through a 30% drawdown without bailing out. No backtest captures this.

Hypothetical performance — the standard disclosure

The performance numbers shown anywhere on this site (in the strategy backtest UI, on the methodology pages, in any future track-record content) are hypothetical. They were computed by applying a defined rule set to historical data; no investor's actual portfolio achieved them.

Hypothetical results suffer from the standard limitations the SEC Marketing Rule disclosures cover:

• They are constructed with the benefit of hindsight — the rule set was designed knowing which rules tend to work on this dataset.
• They do not reflect actual investor decisions: real money flinches, lags, second-guesses.
• They do not reflect material market conditions that affect actual trading: liquidity constraints, position limits, broker outages, tax-loss harvesting, etc.

The act of designing a strategy and then backtesting it is, structurally, a search for rule sets that worked in the past. The risk is that you find rules that match historical noise rather than rules that capture economically real effects. Every backtest on this site lives with that risk.

Regime dependence

A strategy that worked from 2014–2024 may not work from 2024–2034. Examples of regime shifts we've already lived through inside our backtest window:

• 2018 small-cap underperformance. A factor model that loaded heavily on small-cap value would have looked terrible in 2018; the same model from 2020–2021 would have looked spectacular. Pick your start date carefully — and notice that we let you change it.
• 2020 COVID drawdown + recovery. Most factor models broke briefly in March 2020, then recovered as the regime normalized. Strategies that exited on the drawdown locked in losses; strategies that held captured the recovery. Backtest results that span this period silently include or exclude this dynamic depending on rebalance cadence and stop-loss rules.
• 2022 rate regime shift. Long-duration assets repriced sharply when real rates rose. Factor models calibrated on the 2010s low-rate regime degraded materially in 2022. Walk-forward analysis (which we surface on the strategy page) is the only honest way to test for this.

If a backtest looks great on a single 10-year window, that is one observation, not a statistic. We surface walk-forward results — multiple non-overlapping windows — precisely because a strategy that beats SPY in every walk-forward window is qualitatively different from a strategy that beats SPY only on the lucky window.

Factor decay

Factors — value, momentum, quality, low-volatility — go through periods of strong performance and periods of underperformance. In the academic literature this is sometimes called "factor decay" or "factor crowding": once a factor is widely known and traded, its excess return shrinks toward zero. Our 24-factor model is not immune to this; it is calibrated on historical data and has no mechanism to know in advance which factors will continue to work.

Practical implication: do not interpret a high backtest CAGR as a forecast of forward CAGR. The number is the answer to "how would this rule set have performed in the past?" — not "how will this rule set perform in the future?". The two questions have systematically different answers.

Per-strategy disclosure attached

Every backtest result panel inside the application includes:

• The exact date range used
• The benchmark (SPY by default, configurable)
• CAGR, Sharpe, max drawdown, win/loss ratio
• A short caveat block linking back to this page

If you screenshot or share a backtest result anywhere outside this site, the disclosures travel with it as long as the screenshot includes the panel — which it should.

SEC Marketing Rule context

Tessera Research is not a registered investment adviser. Backtest results published on this site are educational research, not personalized investment advice. They are not used to solicit investment advisory clients, and they are not part of an offering of any investment product. The SEC's Marketing Rule (Rule 206(4)-1 under the Investment Advisers Act) governs how registered investment advisers may present hypothetical performance to clients and prospects. While we are not subject to that rule, we voluntarily follow its disclosure framework — labeling hypothetical performance, attaching the relevant disclosures, separating performance claims from solicitation surfaces — because we believe it is the right standard.

If you are an adviser, broker-dealer, or other regulated entity considering republishing or referencing Tessera backtest output in your own marketing, you are responsible for ensuring your usage complies with the rules that apply to you. Email support@tesseraalpha.com if you need attribution language or a citation format.

What this is — and isn't

This page exists so that no number on this site is published without context. We believe the alternative — leading marketing surfaces with backtest CAGR figures and burying the assumptions — is misleading regardless of regulatory posture. If we ever publish a number that contradicts this disclosure, that's a bug, not a marketing decision; please report it.

Read the full methodology →