Corrections log: what changed, why and when

We get things wrong. Tutorial content goes stale especially fast — a platform changes its back end, a library renames a field, a fee gets adjusted. When we're wrong we don't delete the post and pretend it never happened; we log it here in the open: which page, how it originally read, what it changed to, why, and when. You can hold this record against the article and judge for yourself whether we got the fix right.

How we handle a correction

Factual errors that would mislead your actions (such as a wrong field or a parameter stated backwards): we fix the article as soon as possible and leave a record on this page.
Out-of-date information (platform UI, API or fees changed): we update the article and note "follow what the OKX site currently shows."
Pure wording, no effect on understanding (typos, formatting): fixed directly, usually without a separate record.

If you find a piece that's wrong or out of date, write to [email protected], or see the contact page. Corrections we adopt go in the log below.

Log

Date	Page	Original → fix	Reason
2026-10-24	OKX API quant starter, Troubleshooting OKX API errors	The beginner guide had readers create a demo Key, but its balance check code never switched on the demo header → added `set_sandbox_mode(True)` plus a self-check line. The errors guide filed "Key doesn't match the demo / live environment" under 50111 → it now has its own row as 50101 (APIKey does not match current environment), and 50111 covers only an invalid apiKey.	Running the old code as written returns 50101, and the error table pointed readers at a different code. Checked against the error code table and the demo trading section of OKX's API v5 docs.
2026-10-12	OKX Grid Bot Settings	The text said that with the same investment and double the grid count, each grid "earns about half as much in absolute terms" → corrected to: with the range and the amount unchanged, the spacing and the coins per grid both halve, so the gross profit per grid falls to roughly a quarter, with the arithmetic added (10 grids: 2,000 × 0.001 = 2 USDT; 20 grids: 1,000 × 0.0005 ≈ 0.5 USDT).	The old wording left out one factor: doubling the grid count also halves the spacing.
2026-10-11	OKX Funding Account vs Trading Account, OKX vs Binance API, and a number of strategy and API articles	Several articles were rewritten against OKX's API documentation and Help Center, with rules and limits now stated as the official pages give them and linked to their source. Two pages changed title and scope: the former “What Is OKX?” became the account-structure article, covering only the funding account, trading account, sub-accounts and API keys, and the Binance comparison became an API connection comparison → the fee-tier content those two pages used to carry was dropped.	The earlier text made a number of statements without a source; the rewrite keeps only what can be read in the official docs and help pages. The 2026-10-04 fee-tier correction below refers to these two pages as they were before the rewrite.
2026-10-04	OKX Grid Bot Settings	The text described a spot grid as terminated once price rises above the upper limit of the range → corrected to: above the highest price the bot only stops placing orders and has not ended; it ends when take-profit or stop-loss is triggered or when you stop it by hand.	Reading “stops placing orders” as “the bot has ended” leads readers to think the funds have already been settled back to the account. Follows OKX's Help Center description of the spot grid.
2026-10-04	How to Backtest a Strategy	What the text said about how many candles one `fetch_ohlcv` call can pull did not match the API docs → corrected to: the Get candlesticks endpoint returns at most 300 per call.	With the per-call cap wrong, the fetch code as written returns a different number of candles from what the reader expects. Follows OKX's API documentation.
2026-10-04	What Is OKX? (now titled OKX Funding Account vs Trading Account), OKX Fees Explained, OKX vs Binance API (as it read before the rewrite)	The text tied the trading fee tier to OKB holdings → corrected to: the tier is set by asset balance or 30-day trading volume, and holding OKB does not affect it.	The old wording suggested that holding more OKB lowers your fee rate. Follows the fee page on OKX's site.
2026-10-04	OKX Fees Explained	The text described the invite-code fee discount as a smaller fee charged at order time → corrected to: the fee is still charged at your account's tier when the order fills, and the discount is paid back afterwards as cashback to the funding account.	A reader checking fees against their fills would see no reduction and conclude the discount had not applied.
2026-10-04	What Happens When an API Key Leaks, OKX API quant starter, OKX API auto-trading risk control, Troubleshooting OKX API errors	The text was inaccurate about what happens to an idle key with no IP bound → corrected to the rule as documented: an API key that is not bound to an IP and has trade or withdraw permission is deleted automatically after 14 days of inactivity.	Once the key is deleted the script suddenly cannot connect, and readers need to know why and after how long. Follows the API key security section of OKX's API documentation.
2026-09-16	OKX API auto-trading risk control, OKX API quant starter, Troubleshooting OKX API errors	Four points re-checked against OKX's official API documentation. ① We had written that when a custom order ID `clOrdId` repeats, "the exchange only recognises the first order" → corrected to: it deduplicates only among the orders in your account that are not yet in a terminal state; once an order fills or is cancelled the same `clOrdId` can be used again, and a resend is accepted as a brand-new order. The code that looks an order up by `clOrdId` was also switched from the open-order / history list endpoints to the order-details endpoint (those two list endpoints have no `clOrdId` in their request parameters at all). ② We had written that a key "with withdrawal off can't even move funds between accounts" → corrected to: by the official definition, the Trade permission itself includes funding transfers and part of the account settings; what the withdrawal permission governs is on-chain withdrawal. ③ We had written that after cancelling "you get back the same order ID with status canceled" → corrected to: the cancel response carries no status field, it only means the request was accepted, and whether it really cancelled has to be checked with a second query; the code block now includes that step. ④ We had grouped 50120 "insufficient permission" together with 50102 / 50105 / 50110 under "at the HTTP layer they're all 401" → corrected to pull 50120 out on its own: it comes back as HTTP 200 and can only be identified by the code in the response body.	① and ③ make people read "it was sent" as "it was done": resending an order can cause a duplicate fill, and believing a cancel went through when it didn't are both misjudgements that cost money directly. ② stated the permission boundary too narrowly, which could lead readers to underestimate what a leaked key can do. ④ makes error branches written against status codes miss an entire class of permission error. All four follow OKX's official API documentation (checked 2026-09-16).
2026-09-08	How to use the demo account, OKX API quant from scratch	The text said writing `'options': {'sandboxMode': True}` in the config had "the same effect" as `set_sandbox_mode(True)`, and that the live API Key could be reused for the demo → changed to: only `set_sandbox_mode(True)` gets you onto the demo — the demo header `x-simulated-trading: 1` is added by that method itself and writing the value into the constructor's `options` never calls it; and the demo needs its own Demo Trading API Key, created inside the demo environment. We also added that demo and live share one REST host and are separated only by that header, plus a self-check `print` line in both code blocks.	Together these two claims let a reader believe they were on the demo while the request carried no demo flag — and if the Key in hand was a live one with trade permission, the orders going out were real. This time the conclusions were checked against ccxt's own implementation of `set_sandbox_mode` and the demo-trading section of OKX's API docs.

This log will keep growing. You can use it to judge how far to rely on what this site publishes.