The chart above is the entire story compressed into one image. The alpha column only — copy-trade return minus SPY return for that year, per trade. We deliberately don't show absolute returns side-by-side, because "you would have made 5% in 2020" is the same misleading framing the apps use. What matters is the gap. A strategy that returns +5% in a year SPY returned +6% is not a winning strategy — it's a slightly worse, more expensive version of buying the index.
The benchmark we're using on this page is not "safe." It's just transparent.
We compare copy-trading against SPY because SPY is the cheapest, most transparent way for a retail investor to get equity exposure. That doesn't make SPY a safe asset. Two facts a reasonable person should hold in mind while reading the rest of this page:
- The 2023–2026 bull run that powered SPY (and copy-trading alongside it) has been disproportionately driven by a small number of AI- and AI-adjacent megacaps. Index concentration is at multi-decade highs. The top 10 holdings of SPY now make up roughly a third of the entire fund.
- A wave of capital-intensive, unprofitable AI companies — OpenAI, Anthropic, Perplexity, xAI, and others — is approaching the public-market window. Once they list and qualify for index inclusion, they enter SPY at scale. If the AI capex cycle reverses, or if any of these names hits a serious earnings or solvency event, the index takes a write-down disproportionate to the broader economy.
This is not a forecast. It is a structural fact about index composition in 2026. SPY is not the diversified blue-chip proxy it was in 2010. The benchmark on this page is doing well right now for the same reason copy-trading is doing well right now: easy money in mega-cap tech, fueled by an AI capex cycle that may or may not survive its first cyclical test. If that breaks, both correct together — and the diversification SPY appears to provide will turn out to be illusory.
The argument below — that copy-trading delivers SPY-equivalent returns marketed as insider alpha — stands independent of whether SPY itself is overvalued. Both can be true at once. If you are concerned about an AI-bubble correction, neither strategy on this page protects you. Position sizing and asset allocation are separate questions outside the scope of this analysis.
1. 2020: the only year copy-trading meaningfully beat SPY — the COVID-stimulus blowout window. Politicians + Fed liquidity + zero-rate environment.
2. 2021–2022: approximately break-even with SPY. Strategy captures market beta and nothing more.
3. 2023–2024: negative alpha. The strongest bull years of the decade, and copy-trading still trailed simply buying the index.
4. 2025: small positive alpha. Window dominated by Trump-administration policy momentum.
5. 2026 YTD: −4.1% behind SPY. Volatile macro, tariff regime, AI-rotation cycles — copy-trading drifts behind the index.
The strategy is making money right now. So is SPY. The pitch — "trade like Congress, beat the market" — is not what the data delivers.
The 7-year backtest
Strategy: for every disclosed STOCK Act buy disclosure from 2020 through early 2026, the simulated copy-trader enters 45 days after the politician's trade date — at that day's closing price — and exits 90 days later. Each trade is dollar-weighted equally; we then average across all trades opened in a given calendar year. The benchmark is the average rolling 90-day SPY return over the same year.
Below is the result. Universe: 38,468 trades, 164 politicians, options exercises and below-$1,000 transactions excluded.
The chart above is the entire story compressed into one image: the alpha column only — copy-trade return minus SPY return for that year, per trade. We deliberately don't show the absolute returns side-by-side, because "you would have made 5% in 2020" is the same misleading framing the apps use. What matters is the gap, not the raw number. A strategy that returns +5% in a year SPY returned +6% is not a winning strategy. It's a slightly worse version of buying the index.
Why the strategy looks like it works right now
The market is at all-time highs. The Trump administration is pumping fiscal and policy momentum into a narrow set of sectors. Every long-only strategy is making money. That includes copy-trading congressional trades. It also includes throwing darts at the S&P 500 and buying QQQ.
A strategy that makes money in a rising market and loses money in a falling market is not capturing informational edge. It is capturing market beta. SPY is the cheapest, simplest, most transparent way to capture market beta. There is no app fee, no brokerage friction, no 45-day delay, no politician's name attached as a marketing prop.
During the 2020 stimulus pump, this strategy genuinely outperformed (+3.07 percentage points of alpha). During every subsequent year, it has tracked SPY roughly evenly — slightly above in two years, slightly below in three — and the average across the full window comes out negative once you account for any app fee.
What the marketing pitch claims vs. what the data shows
| The claim | What our backtest shows |
|---|---|
| "Trade like Congress, beat the market" | The strategy has matched or modestly underperformed SPY in 6 of the last 7 years. 7-year average alpha before fees: −0.39%. |
| "Politicians have informational edge — copy them" | Politicians may have edge, but most of their disclosed trades are dollar-cost averaging, retirement contributions, spousal trades, index-rebalancing — not informed catalyst plays. Aggregating them washes the alpha out. |
| "Real-time copy of insider trades" | Disclosures arrive 30–45+ days after the actual trade. The platform is operating on data already old enough that the catalyst has typically resolved. |
| "Outperformance shown in our backtest" | Outperformance windows tend to be cherry-picked single years (e.g. one politician's 2020). Across multiple years and the full politician universe, the alpha disappears. |
None of this proves these apps are scams. It proves the marketing pitch overstates what the strategy can structurally deliver.
I interned on the floor of the American Stock Exchange in New York in 2003, when I was 22. I've wanted to build something in this space my entire adult life.
I had the idea for a "copy Congress" app years before any of these platforms existed. By 2021 I had the same dataset everyone else has now. I sat down with the actual STOCK Act data, ran the math, and stopped. The reason is everything you just read above.
The only way to make a copy-Congress app a real consumer business is to sell it as something it isn't — to imply abnormal returns when the data delivers something close to SPY. I couldn't bring myself to do it. I think to look at this same backtest and still ship the marketing pitch, you have to be willing to mislead people who can't easily verify the claim themselves. That's not a product I want to build.
What I built instead is GovGreed. Different product. We surface the layer underneath the disclosure — committee assignments, upcoming markup schedules, line-item bill carveouts, executive-branch financial holdings, federal contract awards, hedge fund 13F overlap, donor mapping, lobbying activity. Forward-looking signal that runs ahead of the 45-day clock. We don't auto-execute. We don't connect to your brokerage. We don't take a fee on your trades. We tell you what's actually moving the price and let you decide whether it's worth acting on.
I think every founder who has run this backtest has reached the same conclusion. Most of them keep selling anyway.
— Stephen Clare · IPS Innovative Platform Solutions · GovGreed.com
The structural reason it cannot deliver alpha at scale
The 45-day STOCK Act window is not a paperwork artifact. It was negotiated by the people whose trades are being disclosed. The window is, by design, longer than the typical move duration of any catalyst that benefits from non-public legislative information. By the time a trade clears the disclosure window:
- Bills that were going to pass have passed (or visibly stalled).
- Federal contract awards have been announced.
- Earnings the politician may have anticipated have been reported.
- Sector rotations triggered by policy news have already happened.
A copy-trader entering on day 46 is buying into the post-catalyst drift, not the catalyst itself. The window is calibrated to ensure the public can never act on the trade in time.
The same logic applies to apps that claim to "copy Warren Buffett." Berkshire Hathaway files its 13F-HR up to 45 days after each quarter ends, and the filing reports positions as of quarter-end, not the day the trade was executed. Effective lag: 45 to 135 days. Plus a complete blind spot for any position opened and closed within a single quarter.
None of this is hidden. It is in the SEC's own documentation. Every founder in this category has seen it.
What we surface instead
The disclosure is the receipt. The cause is what moves stocks. The cause is publicly visible — usually days to weeks before any related disclosure shows up.
- Upcoming committee markups that determine which bills advance and which die. Public schedule. We track 17,585 of them.
- Line-item carveouts in appropriations bills naming specific contractors. Public bill text. We extract these with LLM analysis.
- Federal contract awards winding through USASpending. Public, rarely watched. We track 36,956.
- Federal equity stakes being negotiated (CHIPS Act, Defense Production Act, Loan Programs Office). Sometimes pre-announced in 8-K filings before mainstream coverage catches up.
- Executive-branch personal holdings (OGE 278) showing a Cabinet member personally owns a ticker that overlaps with policy they are about to set.
- Hedge fund 13F overlap with congressional trades — when smart money and Congress are on the same side, that is a signal independent of either disclosure.
None of these are insider information. All of them are public. Almost none of them are surfaced by copy-trading apps because they don't fit the "auto-execute the politician's trade" product shape.
The trade is the receipt. We track the cause.
GovGreed is the intelligence layer underneath congressional disclosures. No auto-execution. No fees on your trades. No brokerage connection. You see the reasoning; you decide whether to act.
Free Account · See the Layer UnderneathMethodology
Every number on this page is reproducible from our public dataset. The query that produced the year-by-year table is approximately:
WITH buys AS (
SELECT id, ticker, trade_date AS pol_d,
trade_date + INTERVAL '45 days' AS copy_d
FROM historical_trades
WHERE LOWER(transaction_type) LIKE '%purchase%'
AND amount_min IS NOT NULL
AND ticker IS NOT NULL
AND trade_date >= '2019-11-01'
AND trade_date < CURRENT_DATE - INTERVAL '135 days'
AND COALESCE(description,'') NOT ILIKE '%exercis%'
)
-- For each buy: enter at copy_d close, exit at copy_d + 90d close
-- Compare to SPY 90-day rolling return averaged over same year
Universe: 38,468 trades after filtering for missing prices, exercises, and trades too recent to have a full 90-day forward window. Daily prices sourced from FMP and Yahoo (1.94M rows). SPY benchmark uses the same daily price database; rolling 90-day return averaged across all trading days in a year. No survivorship bias filtering applied (we use all politicians who traded, not just the ones still in office).
Limitations: this is an aggregate-level test. It does not isolate the small subset of trades that may genuinely carry informational edge. We believe such trades exist — our forward-looking signal layer is built around identifying them before the 45-day clock — but they cannot be reliably extracted by simply auto-copying every disclosed trade.
FAQ
If copy-trading underperforms SPY by less than half a point on average, why call it disingenuous?
The performance is not the issue. The pitch is. "Trade like a senator, beat the market" implies abnormal returns. The math says it returns approximately what the market returns, sometimes worse, before any app fee. Selling market beta dressed up as insider alpha to people who can't easily verify the claim themselves is the part we object to.
Is GovGreed itself a copy-trading platform?
No. We do not connect to a brokerage, do not auto-execute trades, and do not take a fee on your trades. We are an intelligence layer. The user reads the data and decides what, if anything, to do with it.
Does GovGreed have a backtest of its own forward-looking signals?
Yes. Our internal signal-engine backtest shows the A+ tier producing a 72.7% win rate and +10.7% average 30-day return on quarter-deduped trades. That is a different test from the one on this page — it is forward-looking signal, not disclosure-following — and it is calibrated to the asymmetric trade subset rather than the full disclosure firehose.
Where can I see your full data?
The data atlas is at /internal/data-atlas. 218M FEC contributions, 190K STOCK Act trades, 61K bills with text, 7,800 companies, 538 politicians — all cross-referenced. Sixteen public federal data sources. None of it gated behind copy-trading orchestration.
Not financial advice. All data sourced from public federal disclosures (SEC, House Clerk, Senate Secretary, FEC, Congress.gov, OGE, USASpending) and our own derived backtest run on 2026-05-06. We do not reference specific company names because we are describing a category and a mechanism, not making accusations against any individual operator. Operators in this category are free to publish their own backtest in response.