Mean Reversion Dip

Two examples show a complete configuration: the simple example is limited to entry, profit target, stop-loss and risk limits. The example with partial profits matches the template of the dialog and additionally contains a partial-profit plan and the settings for averaging down.

Set up the strategy

The Mean Reversion Dip is assigned under Rule-based Trading to a security of the tree; it is not offered on the portfolio based strategy or on an asset class. The portfolio based strategy must be linked to a watchlist, and the security must be part of that watchlist. Allocate each instrument to exactly one asset class with a weighting. The weightings of the asset classes and those of the securities within each asset class must each total 100%. This makes the asset class of the security tactical: its weighting is then only an upper limit, and a portfolio rebalancing no longer buys into it, not even for its other securities. Preferably put securities with a Mean Reversion Dip in an asset class of their own, see Rebalancing and Mean Reversion Dip together.

In the dialog Strategy definition select Mean Reversion Dip under Strategy name. The YAML editor is pre-filled with the example with partial profits, and Load Template restores it. Keep Active selected for an executable strategy. Clear it to save a draft that still contains settings which cannot be executed yet. Apply saves the strategy. An active strategy is rejected when a required setting is missing or a setting is chosen that cannot be executed yet; the message names the setting concerned.

Effective settings

The following tables describe every setting that influences a decision of the strategy. Shares are written as decimals: 0.03 means 3%, -0.12 means a decline of 12%. Required settings are marked with an asterisk.

The instrument is not selected in the YAML; it is always the security the strategy is assigned to. Data source, execution type and selection mode are fixed: the strategy works on daily closing prices with market execution and without a fee or slippage model of its own. You can leave such settings out. Older configurations that still contain them remain valid.

Name and direction

Cooldowns

Cooldowns only restrict new entries; exits are never postponed.

Entry

Profit taking

The whole profit_management section is optional. Without it the strategy takes no profit and closes a position only through loss or risk rules.

Loss handling

Risk limits

Daily decisions

The strategy uses the last completed trading day before the current UTC day. It skips weekends and recorded exchange holidays; cryptocurrencies use calendar days. A missing expected close or insufficient indicator history produces an unavailable decision. It does not substitute a live quote.

A long entry compares the close with the price T observations ago, the highest prior close in a window, or a prior SMA/EMA. A short entry mirrors the movement and uses the lowest prior close for the window reference. The current close is excluded from these entry references. Short positions require a margin instrument. A margin recommendation also requires an existing transaction with a consistent value per point; different contract definitions for the same instrument make evaluation unavailable.

The hierarchy ceiling, bucket allocation, instrument allocation and position exposure limit all apply; hierarchy weights such as 50 mean 50%. Proposed entries reserve capacity for other strategies in the same evaluation. Short proceeds do not increase equity, and an exit proposal does not release capacity until a trade is recorded. An entry exceeding the remaining capacity is blocked rather than reduced automatically.

Assign actual transactions

When recording a purchase or sale, select its Strategy assignment. Assign both openings and exits consistently, including existing open positions. Multiple strategies may hold the same instrument with separate assignments. Closed, unassigned historical trades do not block evaluation; unassigned open quantities do. A margin closing transaction inherits the assignment of its opening, and changing that opening also updates its closes.

Position quantity, weighted average entry price and cooldown dates are reconstructed from the assigned transactions, with stock splits taken into account. Editing or deleting trades therefore changes the next decision. Removing a strategy leaves the portfolio transactions intact and clears their assignment. Receiving a notification never advances fill state.

Exits and cooldowns

If a risk limit is breached and the forced exit is switched on, a full close is proposed first. The hard stop is evaluated next, followed by the additional loss condition when one is present. Indicator and statistical rules use their configured comparisons; every rule within a group must pass. A hybrid downside condition requires the price threshold and both rule groups. A z-score cannot be calculated from constant prices and is reported as unavailable.

Baseline exits close the strategy’s entire remaining quantity and take precedence over taking partial profit. An exit suppresses a conflicting entry for the same instrument in the same evaluation. Entry cooldowns count calendar days since the last opening or reduction fill. The rolling 30-day trade limit counts actual fill transactions, including partial fills.

Take partial profit

Instead of closing a position only when it reaches its profit target, you can reduce it in several steps. Switch on partial sales in the take-profit section and describe a plan of individual steps. Each step carries its own short name, a condition and the share that is given back when the condition is met. Instead of a share, the last step may close the entire remaining quantity. A long position is sold, a short position is covered.

You choose what the shares refer to. Measured against the opening size, each share is a part of the quantity the position was opened with, and the shares together must not exceed the whole position. Measured against the current size, each share is a part of what is still open when that step fires. Two steps of 30% each therefore give back 30% first and then 30% of what remains.

A step is triggered by a percentage gain over the initial or the weighted average entry price, by a profit amount in the tenant currency, or by indicator rules. Steps are worked through in order and never skipped. When a single price jump reaches several targets at once, the steps concerned are combined into the same evaluation.

Each step executes at most once per position. What counts is only how much of the position has already been given back; receiving a notification never advances fill state. If you record only part of a proposal, the next evaluation proposes exactly the missing quantity again, and restarting the program changes nothing. Never more than what is actually open is proposed. If the position is closed and later reopened, the plan starts over. Executed targets with the same tranche name remain fixed; other progress is reconstructed from the configured plan and recorded reductions.

Add to a losing position

Set loss_action to B_average_down. This setting alone selects averaging down; the settings of the sell-loss variant are then not evaluated. Averaging down requires the additional loss condition under trigger. Long positions add at lower prices; short positions add at higher prices. Every addition requires the adverse-price threshold. Indicator or statistical modes add their respective confirmations; hybrid requires both groups. Missing history never produces an addition.

Under variant_B_average_down you set the size of each addition (add_sizing, as for the entry), the maximum number of additions (max_adds) and the step rule (add_step_rule). With the initial entry as step reference, a 10% threshold and 5% step allow additions at 10%, 15% and 20% adverse movement. Only executed additions advance the ladder. With average cost as reference, each addition requires the configured step from the current weighted average, alongside the downside trigger. Indicator-based steps (indicator_based) use the indicator confirmation without a percentage ladder and require the decision basis indicator or hybrid. Custom steps cannot be activated. The average entry price is recalculated after every addition.

Parent allocations, position exposure limits, cooldowns and the rolling trade limit still apply. Oversized additions are blocked. A drawdown breach blocks additions even without forced liquidation; enabling forced liquidation requests a full exit. Profit-taking and exits have priority over additions, including conflicting recommendations from another strategy on the same instrument.

The first fill of an addition signal consumes one addition; further fills of that signal consume none. Unfilled proposals do not count. Each manual increase without a signal link counts separately. Partial fills of the original entry retain their opening identity. Closing and reopening resets the addition count.

Current-position profit tranches include added units when a new tranche triggers. Its target becomes fixed when the first fill executes; subsequent additions cannot enlarge that started tranche or repeat a completed one. Manual reductions consume outstanding tranches in order, based on holdings before their first reduction. Initial-position tranches retain the opening basis.

Review the result

Open Alert diagnostics from the portfolio’s alert view and select the tab Trading decisions. Which columns it shows, what the actions Buy, Sell, Hold and Blocked mean and when the table stays empty is described in Alert diagnostics. Refresh only reads results, Evaluate now recalculates them. Only Buy and Sell lead to a notification; its delivery is shown in the tab Notifications. The column Reason explains every decision. The following table lists all reasons together with the action under which they appear. With Blocked, some reasons resolve themselves, for example a cooldown; the others require you to complete the configuration or the data.

If the column shows a technical text, the strategy could not be evaluated for another reason; the message is shown unchanged.

How the strategy would have decided in the past is shown by a Historical replay in a simulation environment. There the fee models of the securities accounts apply. Averaging down uses the same daily-close decision and fill interfaces.