Point-in-Time Fundamentals from SEC EDGAR: Five Ways a Backtest Sees the Future
A backtest on fundamentals is only as honest as its dates. Every figure has to be the one that was public on the day the strategy would have traded, and SEC EDGAR data breaks that in five separate ways. Each one below is a real filing, and the last one is a mistake this site made until this week.
1. The period end is not the day you knew
A quarter's numbers become public when the 10-Q is filed, weeks after the quarter closes. Tesla's quarter ended 31 March 2025; its 10-Q was filed on 23 April 2025. A backtest that dates those figures to 31 March trades on them for three weeks before anyone could have. Key every figure by its filing date, not its period end.
2. Restatements rewrite history
Tesla's first-quarter 2025 10-Q reported net income for the first quarter of
2024 as $1,390 million. Its first-quarter 2024 10-Q had said
$1,129 million. The filing tags the difference itself: a
Restatement=ScenarioPreviouslyReported value of $1,129M and a
RestatementAdjustment of $261M. Both numbers are correct. Which one
belongs in a backtest depends on the date: before 23 April 2025 the market knew
$1,129M, after it $1,390M.
A source that serves only the latest figure gives the backtest the revision a year early. The fix is to keep every filing's value and pick the latest one filed on or before the backtest's date.
3. The fourth quarter is never filed
A 10-K reports the fiscal year, not its last quarter, and there is no 10-Q for Q4. Apple's fiscal 2025 revenue was $416.16 billion; its three 10-Qs reported $124.30B, $95.36B and $94.04B. The fourth quarter is the difference: $102.47B, which is the figure Apple announced. Any quarterly series has to derive Q4 this way, and should say that it did.
4. Cash flow in a 10-Q is year-to-date
Income statements in a 10-Q show the three months. Cash-flow statements usually do not: the second-quarter 10-Q shows six months and the third shows nine. Read the nine-month figure as a quarter and operating cash flow triples. The quarter is the nine months minus the six.
5. A year and its fourth quarter end on the same day
This is the one we got wrong. On 26 June 2026 Stryker filed an 8-K recasting its financial statements. It tags fiscal 2024 revenue of $22,595 million and fourth-quarter 2024 revenue of $6,436 million. Both periods end on 31 December 2024. An 8-K carries no fiscal-period label, so a pipeline keyed on (company, figure, period end) sees two values for one key. Ours kept the first in the file, the quarter, and our restatements feed reported that Stryker had cut its 2024 revenue by 71%. It had not.
The fix is to read periodic reports only (10-K, 10-Q and their foreign equivalents, which always say which period they cover) and to key durations by their length as well as their end date. Registration statements and 8-Ks are also where pro forma figures and acquired companies' statements live, which is another reason not to read them as the filer's own results.
Querying around all five
BalanceProof applies all of the above: every figure carries its filing date,
as_of returns only what was filed by that day, Q4 and quarterly
cash flow are derived and labelled as derived, and each period is checked
against its own totals.
curl -H "X-API-Key: YOUR_KEY" \
"https://balanceproof.dev/api/company/TSLA/statements?period=quarterly&as_of=2025-04-01"From Python:
pip install "balanceproof[pandas]"
import balanceproof as bp
client = bp.Client("YOUR_KEY")
df = client.panel(["AAPL", "MSFT", "TSLA"], ["revenue", "net_income"],
period="quarterly", as_of="2025-04-01")
df = df[df.derived.str.len() == 0] # filed figures only, if you preferPoint-in-time queries are on the Pro plan; the rest is on the free tier. Every restatement across all companies is in one feed, and each company page shows its own: Tesla, Apple, Stryker.