A pure token redenomination changes the number of token units and the price per unit in opposite directions, leaving a holder's economic value unchanged before costs and market moves. For comparable charts and returns, express historical prices, balances, supply and base-token volume in one declared unit system. Confirm the conversion ratio and each venue's effective boundary before adjusting data.
A lower price after a split does not make the same economic claim cheaper. A larger balance does not establish a gain. Both observations can be consequences of changing the measuring unit, like measuring the same distance in smaller increments. The harder task is preserving that relationship across a price history assembled from different sources.
Establish what the event actually changes
Polkadot provides a historical example. Its official technical FAQ records the 21 August 2020 redenomination at one old DOT for 100 new DOT. That is a conversion between denominations, rather than evidence that holders suddenly acquired 100 times the economic value. Polkadot launch and redenomination history
Kraken's contemporaneous announcement explains the implementation: one DOT changed from 10¹² Plancks to 10¹⁰ Plancks, while a wallet's Planck balance stayed unchanged. The announcement also said Kraken would begin listing the new denomination on 18 August, ahead of the network's 21 August date, and keep DOT as its trading symbol. Kraken's historical denomination notice
This matters for chart reconciliation. The same ticker and calendar date can describe different units at different venues. A network event timestamp alone cannot tell you which denomination a particular exchange's earlier prices already used.
Start with the documented terms: new units received per old unit, asset identity, preserved rights, effective boundary and any deductions. A pure split is the narrow case where everyone in scope receives the same proportional conversion and nothing else changes economically. A migration that also changes redemption rights or removes balances needs additional accounting.
An approved proposal is also distinct from its execution. The governance timelock and execution guide explains how to keep those stages separate. For chart purposes, establish when the source actually began reporting the new unit, rather than assigning every source the proposal's announced date.
Reconcile the holder and supply together
Consider a hypothetical, unleveraged spot token quoted in USD. One OLD converts into 20 NEW. Assume full conversion, unchanged ownership rights, no fees, no issuance or burning beyond the unit conversion, no cash flows and no change in which balances count as circulating. The names OLD and NEW are worksheet labels, not real tickers.
Define r as NEW units received for one OLD. Here r = 20. Multiply token quantities by r and divide the comparable price by r.
| Measurement | Before, in OLD | Pure translation, in NEW | Later hypothetical NEW close |
|---|---|---|---|
| Holder balance | 37.5 | 750 | 750 |
| USD per token unit | $80.00 | $4.00 | $4.20 |
| Holder's marked value | $3,000 | $3,000 | $3,150 |
| Circulating token units | 12,000,000 | 240,000,000 | 240,000,000 |
| Price × circulating units | $960,000,000 | $960,000,000 | $1,008,000,000 |
The mechanical conversion preserves both products: 37.5 × $80 equals 750 × $4, and 12 million × $80 equals 240 million × $4. The holder's share of circulating supply also remains 0.0003125%. Increasing the displayed supply count is therefore insufficient evidence of dilution.
The later $4.20 close adds a separate market move. The holding gains $150, or 5%, and the circulating market-cap calculation rises by the same percentage under these fixed-supply assumptions. Nothing in the conversion guarantees that later price.
Comparing $4.20 directly with $80 produces −94.75%. Those prices measure different-sized token units. With no intervening flows or costs, the comparable holding return is (r × new price ÷ old price) − 1. Here, (20 × $4.20 ÷ $80) − 1 is 5%.
The method also handles consolidation. If 20 OLD instead become one NEW, r is 0.05. The same 37.5 OLD would translate into 1.875 NEW and an $80 reference price into $1,600; their product remains $3,000. Write the ratio in words before calculating, because “20:1” alone can invite the opposite interpretation.
Keep the supply definition fixed while checking these products. The free-float versus circulating-supply comparison explains why provider exclusions can differ. A unit conversion and a revised classification are separate changes; dividing price cannot repair an unexplained change in the counted ownership groups.
Put every candle field on the same basis
For a history expressed in NEW, translate pre-event OLD observations to the NEW basis. Leave observations already expressed in NEW unchanged. Under the hypothetical 20-for-1 conversion, each old price is divided by 20 and each old base-token volume is multiplied by 20.
Applying one factor to open, high, low and close preserves their ordering. Adjusting only the close leaves an inconsistent candle: an apparent high might be quoted in OLD while its close is quoted in NEW.
The following bars are entirely hypothetical. The pre-event and next bars each contain trades in only one denomination. All prices are USD per unit named by the column; volume counts that column's token units.
| Field | Pre-event raw OLD bar | Same bar on NEW basis | Next raw NEW bar |
|---|---|---|---|
| Open | 76.00 | 3.80 | 4.00 |
| High | 84.00 | 4.20 | 4.35 |
| Low | 74.00 | 3.70 | 3.95 |
| Close | 80.00 | 4.00 | 4.20 |
| Base-token volume | 120,000 | 2,400,000 | 2,600,000 |
The next close is 5% above the adjusted previous close. Comparable base volume grows by 8.33%, from 2.4 million to 2.6 million NEW. Comparing 2.6 million NEW with 120,000 OLD instead mixes units and invents a much larger activity jump.
This reciprocal treatment has a documented analogue in equities. J-Quants applies its cumulative adjustment factor to all OHLC prices and the inverse split factor to volume. Those are that equity API's conventions; they do not establish how any crypto feed handles token redenominations. The token example above follows directly from its declared unit conversion. J-Quants adjusted-price methodology
Quote volume needs different treatment. A hypothetical execution of 120 OLD at $79 has $9,480 of turnover. Expressed as 2,400 NEW at $3.95, it still has $9,480 of turnover. Do not multiply USD turnover or the number of executions by 20. Obtain exact quote turnover from the constituent trades; candle close multiplied by aggregate base volume is not generally their exact traded value.
An interval containing both denominations requires finer data. Translate each constituent trade to the common unit before calculating the interval's open, high, low, close and volume. A single factor applied to an already mixed candle cannot tell which extrema belong to which denomination. If the required trades or smaller bars are unavailable, flag that interval instead of manufacturing precision.
Catch adjustment errors before trusting the return
The most dangerous repair is applying a valid factor twice. Suppose a provider has already translated the old $80 close to $4. Dividing its $4 observation by 20 again produces $0.20. The next $4.20 close then appears to gain 2,000%, rather than the correct 5%. A dramatic “correction” can be as false as the original discontinuity.
Inspect the source's documented basis and a few observations on both sides of the boundary. A smooth chart is supporting evidence, not proof of the adjustment method. Real market prices can gap during an event, and an unrelated price move could resemble a split ratio.
Provider labels require care too. StockCharts explicitly says that even its stock charts called unadjusted are corrected for splits; their distinction concerns other adjustments. It also warns that adjusted prices cannot establish the actual historical execution price. This illustrates why a generic adjusted/unadjusted switch is insufficient evidence about a crypto dataset. StockCharts price-adjustment conventions
Retain raw observations beside a derived comparable series. The adjusted $4 close in the example means “the old close expressed per NEW”; it does not mean a NEW-denominated trade occurred at $4 in that earlier market. Reconcile actual orders against the original venue records. The DCA symbol migration procedure addresses the separate task of translating inventory and rebuilding working instructions.
Finally, check whether the event was only a denomination change. A stablecoin's new unit can be perfectly reconciled while its redemption arrangements or backing have changed. The stablecoin collateral comparison helps identify those separate economic questions. Unit arithmetic alone cannot establish equivalent collateral, liquidity or redemption access.
Keep a chart receipt that another reader can reproduce
Save an event record with the dataset, rather than relying on a note attached to a screenshot.
| Receipt field | What to record |
|---|---|
| Instrument | Network or contract, venue, market and quote currency |
| Conversion | OLD and NEW definitions; NEW units per OLD written explicitly |
| Boundary | Source-specific timestamp, timezone and supporting announcement or data evidence |
| Input basis | Raw, already adjusted or unresolved, separately for each field |
| Transformation | Price factor, base-volume factor and fields left unchanged |
| Continuity checks | Holder value, supply × price, quote turnover and sample OHLC ordering |
| Exceptions | Mixed intervals, missing data, costs, supply revisions or altered rights |
| Dataset version | Retrieval date, provider revision and saved adjustment-event table |
For a backtest, declare the unit basis at each simulated decision time. A history in today's units reflects subsequent redenominations; a point-in-time reconstruction uses only events effective and known by that decision time, applied consistently across its lookback window. Older observations in the same window still need a common unit basis. J-Quants, for example, documents retroactive revisions to its adjusted equity data. A comparable crypto research process should retain its own input version and event factors rather than assume today's downloaded history is unchanged. J-Quants revision and point-in-time explanation
Percentage comparisons within a consistently scaled segment survive a pure change of units. Absolute price thresholds, token quantities and volume thresholds must carry the appropriate unit. Recompute indicators from consistent candles, and preserve any genuinely missing trading interval rather than filling a venue pause with invented activity.
For the worked history, the receipt should say exactly what was done: pre-event OLD prices divided by 20, pre-event base volume multiplied by 20, USD turnover unchanged, and later observations already in NEW left untouched. Only after those checks does the move from an adjusted $4 close to $4.20 describe a 5% return.