What it is and when it applies
This page helps you decide whether to buy after a hard fall, only once a confirmation you wrote down has printed, or to pass. An oversold reading is where the plan starts looking, and it is never the buy. The stop goes below the low of the fall, and that distance sets the size.
Oversold here means RSI with its published default, RSI(14), closing under the 30 band. A reading of 30 says the average gain over the window is 0.43 times the average loss. It describes the candles already printed and does not say what the next one will do. In a long decline RSI can sit under 30 for many candles while price keeps falling, so the first touch is a place to start watching, not a trigger.
The belief behind the trade is that a one-sided fall leaves fewer willing sellers, so price can bounce toward the last breakdown level. It is a belief, and this page cites no test of bounces after oversold readings. A fall can run straight through.
It fits a decline that still shows lower highs you can mark, on a coin whose book your own order will not move. It does not fit a vertical collapse with no pause, which gives no level to confirm against, or a thin coin.
Before you enter
The long is wrong if a candle closes below the lowest low of the fall. The decline is then still running. The fall runs from the last swing high before RSI(14) first closed under 30 to its lowest low so far, and that low is frozen at the confirmation close. The stop order goes at the low minus a buffer, below it and not inside the range, because a first bounce can retest it while the idea is alive. The examples assume a buffer of 0.5 x ATR(14), an assumed example and not a rule. A wider buffer takes a bigger move to reach and shrinks the position, and a tighter one does the reverse. Stop-loss placement explains the level, the close and the order. A confirmation that prints well above the low sits far from this stop, so the size is small. That is the price of waiting.
This manual's house constraints are a planned loss of 1-2% of the account per trade and a 3x maximum leverage. They are choices, not a recommended allocation, and less, or no position, is always allowed. Sizing inputs:
- Account risk: the planned loss in cash, 1-2% of equity or lower.
- Entry price: the assumed fill, near the confirmation close.
- Stop price: the fall's lowest low minus the buffer.
- Cost allowance: a per-unit amount for fees and slippage, added to the stop distance, plus funding for the planned hold. The examples assume 0.3 per unit, an illustrative figure (trading costs and liquidity covers it).
- Leverage cap: the leverage setting, and all open position values plus this one over account equity, 3 or less.
- Liquidation distance: the ticket's estimated liquidation price, below the stop for a long. If the gap is not wide, lower the leverage setting.
Units = planned loss / (entry minus stop, plus allowance), rounded down (position sizing has the sum). Units x (distance + allowance) is the planned loss if the stop fills at its price. A gap, slippage or funding can make the realised loss larger.
For arithmetic only, assume a 10,000 account, a planned loss of 1%, or 100, a lowest low of 100 and ATR(14) of 4. The buffer is 2, so the stop is 98. A confirmation close at 109 is 11 from the stop, 11.3 with the allowance. The size is 100 / 11.3, about 8.85, so 8 units, worth 872, or 0.09 of the account. The planned loss with costs is 8 x 11.3 = 90.4, and costs are about 3% of it.
Conditions
Each of these must be checkable on the chart.
- RSI(14) has closed under 30 on the decision timeframe at least once in the last 10 candles before the trigger (assume 10).
- The lowest low of the fall is written down, and so is L, the most recent swing high inside the fall by a swing rule you fixed, before the trigger. A fall with no lower high gives no L and no plan.
- The confirmation is a decision-timeframe close above L. The plain close is the usual convention, and a wick above L does not count (breakouts and retests covers the rule). Your written choice can add a regular bullish divergence, a lower low in price with a higher low in RSI(14), before that close. A divergence is only known after the second swing has turned, and the label can vanish.
- Funding is context, never a signal. Read the contract's rate and history, and add expected payments for the planned hold to the allowance when the rate points against a long. A rate that pays longs is small next to a move against you (funding rates works the sum). This page cites no test of funding against later price.
- There is room: the reward to the first resistance, after costs, passes a test you wrote before the trade. This manual sets no minimum ratio. Risk to reward gives the break-even share to compare against.
Reasons to pass
Passing is a valid result. Pass when:
- RSI is under 30 but no close above L has printed. Wait
- price makes a lower low before the confirmation, which voids the plan: write a new one from the new low, with its own stop and size
- the fall is vertical and there is no lower high to mark
- the size the stop allows is below the exchange minimum, or costs, funding included, come to more than one fifth of the planned loss
- the reward to the first resistance fails your written test
- the book near the low is thin enough that your own order would move it
- other open longs in coins that move with this one already fill the group cap in portfolio risk
Entry
Entry on the confirmation close
The trigger is the confirmation candle's close. Place a limit order at or just below it, or a market order if getting in matters more than the fill price. A limit order expires after 3 candles (assumed). The stop is a reduce-only stop-market sell at the lowest low minus the buffer. A stop-limit with its limit below the trigger caps the fill price but can sit unfilled in a gap (order types explains).
Entry on the retest of L
After the confirmation, price may dip back. Inside a window you fixed (assume 10 candles), with no close below the lowest low, a candle trades down to within a distance you fixed of L (assume 0.5 x ATR(14)), or through it, and closes back above L. Buy near that close. The stop stays at the same flat price. If that close is below the confirmation entry, the distance is shorter and the position larger for the same planned loss. A retest may never come, and one that closes below the lowest low means the bounce failed. Cancel when the window ends.
Execution assumptions
Assume a limit order that fills at its price, 0.3 per unit for fees and slippage, and a stop-market that fills at the next available price. If price runs away without a fill, do not chase, because the stop does not move and the reward shrinks.
Hypothetical walkthrough
Hypothetical walkthrough. Invented prices and an invented outcome, chosen to teach the decision. Not a recorded trade, and not evidence of how often this setup works.
The reader writes the plan when RSI(14) first closes at 24 with price at 100: no order yet. The plan reads: if a 4-hour candle closes above L, the last lower high at 108, with the lowest low still 100, buy near that close with the stop at 98. A close at 109 gives 8 units and a planned loss with costs of 90.4. With the first resistance at 124, the reward after costs is 14.7 against 11.3, about 1:1.3 if every unit closes there.
Now change one fact. Before any close above 108, price falls to a lower low at 96 with no new swing high by the reader's swing rule, so L is still 108. The plan is void and the reader rewrites it from that low: stop 94. The same close at 109 is 15.3 away with the allowance, so 6 units, and the reward after costs is 14.7 against 15.3, about 1:0.96. The reader's own test, written beforehand and not a rule of this manual, says pass if the reward after costs is smaller than the risk after costs. The pass is logged with its reason. The outcome was chosen to show a clean confirmation ending in a pass because the fall had not finished.
Exits
Stop. Keep the reduce-only stop at the lowest low minus the buffer. A decision-timeframe close below that low means the idea is wrong: close by hand and cancel any unfilled orders. The order is the backstop (stop-loss placement).
Taking profit. The first target is the next resistance above, such as where the fall gathered speed, found by looking left and marked before entry. A bounce can end there. Close part and write the rest of the plan before the order (taking profit covers partial exits and the R arithmetic). After a partial, resize the stop to the units left, and once the first target fills, the stop can move to entry or just below the latest higher low. It only moves toward price.
Trailing the rest. The default here is a target. The Chandelier Exit trend ride is optional, for a plan with no fixed target or a part left running, and it takes over only once its line passes the first stop in your favour. Leave it off while the chart still shows lower highs, in a range, or when you cannot watch each close.
Common pitfalls
- Buying the first touch. An oversold reading is a reason to watch. The close above L is the trigger.
- Averaging down while RSI stays under 30. An added unit is a new plan with its own stop and size.
- Treating several oversold coins as separate trades. Coins that fell together can stop out together, so add their planned losses.
- Moving L after the fact. A level redrawn to fit the candle that closed above it can never say the idea is early.
What this cannot tell you
- Whether the bounce is a recovery or a pause. A close above L is one move inside whichever the decline turns out to be.
- Whether RSI under 30 says anything about the next candle. It describes closes that have printed, and this page cites no test of it on perpetuals.
- Where the stop fills. After a fall the book can be thin, and a stop-market fills at the next available price, so the realised loss can be larger.
Questions
Is RSI under 30 a buy signal?
No. It says recent losses were large next to recent gains. Price can keep falling with RSI under 30, so this page buys only after a written confirmation.
Is a negative funding rate a reason to buy?
No. It shows shorts paying longs to hold, which is a cost question. It does not say who will be right.