Grid Fisher Transform EA
- 专家
- 版本: 3.25
- 激活: 5
For grid traders who want to start a ladder at a turning point, not in the middle of a move. One job: open a grid only when price snaps back from a real stretch with the trend behind it. Mechanism: an Ehlers Fisher Transform zero cross on a closed bar, a Stretch Guard, a Range Guard and an EMA 200, then an ATR-spaced grid that banks small closures.
Tested on: EURUSD H1, $10,000, default inputs, 2025.01.01 to 2026.09.30: net +593.99, maximal equity drawdown 177.95 (1.78%), 245 trades, no stop-out, no margin refusal. Deepest basket: 8 grid orders open at the same moment, 0.44 lots in total.
Minimum deposit, from the deepest basket: EURUSD H1: the deepest basket (0.44 lots) needed about $480 of margin at 1:100 on top of a 177.95 drawdown, so start with at least $2,000 at the default 0.01 first lot. XAUUSD D1: the deepest basket was 1 order of 0.01 lots, about $35 of margin with a 32.01 drawdown, so start with at least $500. These are floors from one test, not a safe zone; deposits below $10,000 were not run.
Try it free first: load Grid Fisher Transform EA in the MetaTrader 5 Strategy Tester with visual mode on, default inputs, EURUSD H1. Every crossing it accepts gets an arrow; every crossing a guard refuses is shown on the live chart as a grey dot.
The trade lifecycle in four verbs. It stretches the last 10 bars' price position into a sharp Fisher wave, waits until that wave has reached an extreme and snaps back across zero with the trend behind it, opens the first level and spaces the next ones by ATR, and banks groups of orders as soon as their joint result clears the target.
Plain risk up front. This IS a grid. It adds to losing positions and uses no stop loss by default. Defaults: 0.01 first lot, additive steps of 0.01, ATR spacing with a 50 point floor, hedge cycle off, margin-level safety off. Limits stop new levels; they never force-close the book.
----------
TURNING POINTS, MADE SHARP
John Ehlers' Fisher Transform takes where the bar's median price sits inside its recent high-low range and reshapes that position so that the turns become sharp, clear swings instead of soft rounded tops. Near the top of the range the line shoots up, near the bottom it dives, and when it crosses zero the balance between buyers and sellers has flipped. The EA adds a 5 bar EMA of the Fisher line as a signal line for an optional faster mode.
Five things to know first
• Entry: the Fisher line crosses above zero (BUY) or below zero (SELL) on a closed bar. A signal-line mode is included for earlier, more frequent entries. Accepted signals do not repaint.
• Calculated inside the EA from bar highs and lows. No indicator file, no DLL, no web request.
• The new Stretch Guard only accepts a cross that comes out of a real extreme, a Range Guard skips quiet boxes, and an EMA 200 filter keeps entries on the side of the main trend.
• The Stretch Guard ships switched ON at 1.0, because in our test it cut the worst EURUSD drawdown by 78% and more than doubled profit per unit of drawdown. The full on/off numbers are below, including the profit it gave up.
• No stop loss and no forced close-all. Limits stop new levels; the open book stays under management.
----------
Start here (5 minutes)
1. Attach it to EURUSD H1 and leave every input at its default.
2. Set RunMode to Scan only for the first session. The panel shows each plan; no order is sent.
3. Look at the arrows and grey dots on the chart: they show which past crossings the guards would have taken or refused.
4. Check your balance against the minimum-deposit line above, then press ARMED on the panel.
5. Repeat the tester runs below on your own broker's data before any real money; spread, swap and commission differ between brokers. More than one chart? Give each copy its own EAMagicNumber and HedgeMagicNumber.
----------
Reproduce our numbers
Strategy Tester, default inputs, 1 minute OHLC, leverage 1:100, 2025.01.01 to 2026.09.30, one broker's demo data. Same settings, same period, and you should land close to these:
• EURUSD H1, $10,000: net +593.99, maximal equity drawdown 177.95 (1.78%), 245 trades, deepest basket 8 orders / 0.44 lots. No stop-out, no margin refusal.
• EURUSD H1, $1,000,000: net +593.99, maximal equity drawdown 177.95 (0.02%), 245 trades, deepest basket 8 orders / 0.44 lots. No stop-out, no margin refusal.
• XAUUSD D1, $10,000: net +24.25, maximal equity drawdown 32.01 (0.32%), 7 trades, deepest basket 1 order / 0.01 lots. No stop-out, no margin refusal.
• XAUUSD D1, $1,000,000: net +24.25, maximal equity drawdown 32.01 (0.00%), 7 trades, deepest basket 1 order / 0.01 lots. No stop-out, no margin refusal.
Read these as one past stretch, not a forecast. The deepest basket is counted from the tester's deal list (largest number of grid orders open together, and their lots). Money drawdown is the same at both deposits because lots do not grow with the balance, so it is a far bigger share of a small account. Defaults were set once by judgement on this same period, which flatters them. Seven gold trades in 21 months is far too few to judge gold. Netting accounts were not verified.
----------
The Stretch Guard: on or off, in numbers
The idea: a Fisher cross that only wiggles around zero says little; one that comes back from a deep extreme marks a real turn. So for a BUY the Fisher line must have dropped to at least -InpFTStretchMin in the 10 bars before the cross (for a SELL, risen to +InpFTStretchMin). Same period, same settings, guard at 1.0 against guard off:
• EURUSD H1, $10,000: on, net +593.99, drawdown 177.95 (1.78%), 245 trades, deepest 8 orders / 0.44 lots. Off, net +1,161.91, drawdown 804.04 (8.01%), 458 trades, deepest 11 orders / 0.95 lots.
• XAUUSD D1, $10,000: on and off identical, net +24.25, drawdown 32.01, 7 trades, deepest 1 order / 0.01 lots.
On EURUSD the guard cut the worst drawdown by 78%, more than halved the deepest basket's lots, and raised profit per unit of drawdown from about 1.4 to about 3.3, at the cost of 49% of the profit. On daily gold every cross already came out of a stretch, so the guard changed nothing there. If you prefer the bigger profit and accept about four times the drawdown, set InpFTStretchMin to 0.
----------
Why the Fisher Transform
A grid needs one clean starting gun, then it needs the engine to work. The Fisher Transform is a good gun for three reasons. It is built from price position only, so it means the same on every broker. It is bounded by its own range, so the zero line and the stretch levels work on any symbol without retuning. And it turns sharply, so the cross is a single, clear event on a closed bar. The honest costs: it reacts to the last 10 bars only, so in a strong one-way trend it can flip back and forth near zero, it arrives after the turn has started, and an extreme reading can stay extreme for a while. The guards, the trend filter and the signal gap are there for exactly that.
----------
One cycle, step by step
1. Measure. Each new bar recomputes the Fisher line over 10 finished bars and its 5 bar signal EMA.
2. Cross. In the default mode the Fisher line must cross zero, upward for a BUY, downward for a SELL. In signal mode it must cross its signal line instead.
3. Stretch Guard. In the 10 bars before the cross the Fisher line must have reached -1.0 for a BUY or +1.0 for a SELL (0 switches it off).
4. Range Guard. The last 20 bars must span at least 5 ATR, so tight boxes are skipped.
5. Trend filter. By default the close must be on the matching side of the EMA 200. It can also fade that side, or be switched off.
6. Timing. Accepted signals are kept apart by a gap of bars, and a signal may wait a few bars if the level cannot be placed at once.
7. Ladder. Levels follow ATR or fixed points, with a floor, a ceiling and an optional freeze. BUY and SELL grids can run together or alone.
8. Banking. Groups of oldest and newest orders close when their joint result, after swap and commission, clears the target. If the oldest order is too heavy, a slice of it is closed instead.
9. Safety checks. A closure that would reopen a level at once is blocked, each leg is confirmed in deal history, failed legs are retried, and a plan is dropped if the book changed before sending.
10. Hedge cycle, optional. Off by default and not exercised in our tests. When on, it holds a balanced pair on a chosen symbol and retires it in equal slices, paid by a small surcharge on closures the grid already makes.
----------
The panel
The panel sits on the left, so the right side and the price scale stay clear. It lists symbol, magic number, a status pill, grid mode, lot step, spacing, banked result, the book, each side's closing plan and the order table. The FISHER WAVE row shows the period and signal length, the Fisher and signal values of the last closed bar, whether the line is rising (UP) or falling (DN), and the bars since the last crossing. The STRETCH GUARD row shows the side Fisher favours, the mode, how deep the stretch before it went with OPEN, HOLD or off, the Range Guard width in ATR with its state, and the EMA side. On the live chart every crossing of the last 150 bars is marked: a small arrow when accepted, a grey dot when a guard refused it. Buttons switch Armed / Scan only, Pause / Resume, chart marks and slicing. The Hedge Vault card docks under the main panel; with the hedge off it stays hidden. Nothing is drawn in a non-visual backtest.
----------
Limits you control
No setting closes the book for you; limits only stop new levels.
Each order passes a free margin check with a safety buffer before it is sent.
Caps on levels per grid, orders per side and lot per order keep the ladder finite. MinMarginLevelPercent and MaxAccountRiskPercent start at off; set them for your own account.
An optional margin-level timeframe safety moves the signal and spacing to a higher timeframe as margin falls. It is off by default and was not exercised in our tests.
Copies are kept apart by symbol and magic number.
Scan only shows every decision without sending a single order.
----------
Who this is for
Traders who want a grid that starts when price turns back from an extreme, not halfway through a move; who are happy with many small closures; who prefer a quiet EA that trades less with the trend behind it; and who will size lots to their balance.
Who this is NOT for
• Anyone who needs a stop loss on every order. This grid has none by design.
• Anyone expecting a fixed monthly return or a fast doubling. The tested result above is modest, and it is the honest one.
• Accounts below the minimum-deposit line above. The deepest basket needs room for its margin and its drawdown at the same time.
• Small accounts running large lots unattended. A long one-way trend can hold the book open for weeks.
• Traders who want to trade every candle. The filters keep it quiet on purpose, and on daily gold it may sit idle for months.
• Traders who want the largest possible profit at any drawdown. The default favours a smaller drawdown; switching the Stretch Guard off trades more and swings harder.
• Traders who want a ready-made gold system. Gold was tested only on D1 with seven trades, which proves nothing either way.
• Traders who will not test first. If you skip the Strategy Tester on your own broker, this is not the right tool for you.
----------
Common questions
Does it repaint? Accepted signals, chart marks and both panel rows use finished bars only.
Is it the classic Fisher Transform? Yes, Ehlers' formula: the median price's position in the 10 bar high-low range is smoothed (0.66 new, 0.67 old), clamped below 1, and transformed with 0.5 x ln((1 + v) / (1 - v)) plus half of the previous value. The signal line is a 5 bar EMA of it, added by this EA.
Zero or signal-line mode? Zero mode waits for the balance itself to flip and trades less. Signal-line mode reacts to the Fisher line turning and trades more often; it was not backtested here.
Which markets were tested? EURUSD H1 and XAUUSD D1. Everything else is for you to test.
How deep did the grid go in the test? At most 8 orders open together on EURUSD, 0.44 lots in total, with the default guard. With the guard off it reached 11 orders and 0.95 lots.
How much should I start with? See the minimum-deposit line near the top: at least $2,000 for EURUSD H1 and $500 for XAUUSD D1 at the default first lot, more if you raise the lots.
What should I change first? InpFTStretchMin, then InpFTRangeMinATR, then InpFTPeriod. One change at a time, and look at drawdown and basket depth next to net result.
Netting accounts? A separate margin cap exists for them, but netting behaviour is unverified.
----------
All inputs
Grid levels
TradeMode - two grids, BUY and SELL, by default, or one direction only.
GridSpacePoints - level distance in fixed spacing mode (25).
GridSpaceMultiplier - above 1.0 widens deeper levels (1.0 = even).
MaxGridLevels - most levels in one grid (1000).
UpDirectionIsBuy - which side an up move arms.
EAMagicNumber - number on grid orders (9769330).
GridOrderComment - text on grid orders (GFTR).
OneOrderPerBar - at most one new level per bar.
Lots
InitialLotSize - volume of the first level (0.01).
LotSizingMode - additive by default, multiply as the alternative.
LotMultiplier - factor per level in multiply mode (1.25).
LotAddStep - step per level in additive mode (0.01; 0 uses the starting lot).
MaxOrderVolume - largest single order (1.0).
UseAutoLot - links the starting lot to balance (off).
AutoLotBaseBalance - balance that matches the starting lot (1000).
AutoLotAllowDecrease - lets auto lot go below the starting lot (off).
ATR spacing
GridSpacingMode - ATR by default, fixed points as the alternative.
ATRTimeframe - ATR timeframe (current chart).
ATRPeriod - ATR length (14).
ATRSpacingFactor - spacing as a multiple of ATR (1.0).
ATRBarShift - ATR bar; 1 is the last finished bar.
ATRMinSpacingPts - lowest spacing in points (50; 0 none).
ATRMaxSpacingPts - highest spacing in points (5000; 0 none).
ATRFreezeOnStart - keeps one spacing for the whole cycle (off).
Profit target
DesiredProfitPerClosure - base target in account currency (2.0).
ProfitMultiplier - final scale, 1.00 to 2.00.
ProfitTargetMode - scaled by default: bigger, deeper books aim higher.
ProfitVolumePower - how closing volume lifts the target (1.0; 0 off).
ProfitDepthBonus - extra target per additional open level (0.10).
ProfitMaxMultiple - scaled target ceiling in multiples of the base (50).
ProfitMaxReachFraction - keeps the target inside this share of one grid step (0.50).
Fragment scanner
UseMultiLegScan - checks multi-order groups before the plain pair (on).
MaxOldestLegs - most old orders in one closure (3).
MaxNewestLegs - most new orders in one closure (6).
ScanPriority - ranking of qualifying groups; largest group first by default.
NeverBankLessThanPair - a group must also beat the pair target (on).
RequireCloseVolumeExceedsTarget - newest volume must be above oldest volume (on).
MinCloseVolumeRatio - required ratio of those volumes (1.5).
Target slicing
EnableTargetSlicing - allows part-closing the oldest order (on).
FragProfitWeightPower - profit weight per closing lot (1.5).
FragMinSlicePercent - smallest slice, percent of the target order (20).
FragAllowMultiCloser - several new orders may fund one slice (on).
FragPreferFullClosure - a full closure wins when both are possible (on).
Reopen guard
ReopenGuard - blocks closures that would reopen a level at once (step based).
ReopenGuardCushion - safe distance as a share of one step (0.90).
Account protection
MinMarginLevelPercent - no new levels below this margin level (0 off).
MaxAccountRiskPercent - no new levels beyond this floating loss, percent of balance (0 off).
NetMaxMarginPercent - netting only: cap on net position margin, percent of equity (5.0).
MaxTotalOrders - absolute order cap per side (1000).
Execution and checks
RunMode - Armed or Scan only.
CloseSequence - order in which legs are sent; oldest first by default.
MaxSlippagePoints - allowed slippage (30).
MaxSpreadPoints - no trading above this spread (7500).
MinSecondsBetweenClosures - pause between closures per side (5).
CloseAttemptsPerLeg - attempts before the retry queue (3).
VerifyTimeoutSeconds - how long a stuck leg is followed (90).
DeferWhenMarketClosed - waits for the session instead of halting (on).
MarketClosedRetrySeconds - retry spacing while orders are refused (30).
AutoResumeAfterSeconds - a halt clears itself after this many seconds (300; 0 waits for you).
NoPartialRetryMinutes - partial closes tried again this many minutes after a broker refusal (15; 0 never).
AbortIfBookChanged - drops a plan if a leg changed (on).
UseFullCommission - counts commission on both sides (on).
CommissionPerLot - fallback figure (0 reads history).
CloseOrderComment - text on closing deals.
Hedge cycle
EnableHedgeCycle - main switch (off).
HedgeSymbols - comma list; repeat a symbol for more turns (XAGUSD).
HedgeSymbolRotation - how the next symbol is picked (per cycle).
HedgeMaxSymbolsInTester - rotation cap in the tester only (2; 0 none).
HedgeCycleLots - lots on each side of the pair (0.01).
HedgeSliceLots - lots retired per side each slice (0.01).
HedgeFragmentBonus - bonus over the slice cost (100).
HedgeBonusIsPercent - bonus is a percent; false means money (on).
HedgeMaxSurcharge - money cap per closure (0 none).
HedgeMaxSurchargePercent - largest rise of any closure target, percent (50).
HedgeMagicNumber - number on hedge legs, must differ from the grid (9769331).
HedgeReopenOnFinish - opens a new cycle when one ends (on).
HedgeStartRequiresTrade - first cycle waits for a grid order (on).
HedgeIntegritySeconds - leg balance check interval (5).
HedgeAutoRepair - repairs a lone leg automatically (on).
HedgeNakedAction - what to do with a lone leg; re-hedge by default.
HedgeMaxSpreadPoints - hedge spread limit (0 off).
HedgeSlippagePoints - hedge slippage allowance (30).
HedgeAttemptsPerLeg - attempts per hedge leg (3).
HedgeVerifyTimeout - follow time for a stuck hedge leg (90).
HedgeCommissionPerLot - fallback hedge commission (0 reads history).
HedgeOrderComment - text on hedge deals.
Weekly cash day
WeeklyCashDay - pauses new cycles and slices around withdrawals (off).
CashDayOfWeek - 0 Sunday, 1 Monday and so on (1).
CashCutoffHour - server hour after which no new slice fires (21).
Panel and chart
ShowPanel - panel on or off.
PanelX and PanelY - distance from the chart edge (12, 15).
PanelFontSize - base text size (9).
ShowArabic - second language line (on).
ShowChartMarks - plans and levels drawn on the chart (on).
MaxTableRows - order rows per grid (13).
PanelCorner - corner the panel is anchored to.
Entry trigger
EntryTriggerMode - the Fisher Transform strategy by default.
SignalBarShift - 1 is the last finished bar.
SignalBarsDelay - bars between accepted signals (30; 0 every).
SignalValidBars - bars a signal stays usable (3).
SignalAllowBuy and SignalAllowSell - switch each side.
SignalReverse - turns the direction around (off).
Fisher Transform
InpFTPeriod - bars in the highest high / lowest low window (10).
InpFTSignal - EMA length of the signal line (5).
InpFTMode - 0 Fisher crosses zero, 1 Fisher crosses its signal line (0).
InpFTStretchMin - Stretch Guard: the Fisher line must reach minus this (BUY) or plus this (SELL) in the 10 bars before the cross (1.0; 0 off; see the test above).
InpFTRangeMinATR - Range Guard: least window width in ATR (5; 0 off).
InpFTRangeBars - Range Guard: window length in bars (20).
InpFTATRPeriod - ATR length for the Range Guard (14).
InpFTTrendFilter - 1 with the EMA side, 2 against it, 0 off (1).
InpFTTrendMAPeriod - trend EMA length (200).
InpShowSignalArrows - arrow where a cycle armed (on).
InpShowFTOnChart - past crossings on the live chart, arrow or grey dot (on).
Debug
DebugMode - extra journal detail (off).
Margin timeframe safety
UseMarginTFSafety - main switch (off).
Level1MarginPct and Level1TF - at 6000 percent the work moves to M15.
Level2MarginPct and Level2TF - at 5000 percent the work moves to H1.
MarginTFHysteresisPct - recovery needed before stepping back (500).
----------
Risk, plainly
Grid trading is risky. It adds to losing positions, uses no stop loss, and a long trend can keep the book open for a long time or, with lots too big for the balance, end in a stop-out and a heavy loss. The Fisher Transform is built from past highs and lows and cannot see the next bar. No profit is promised. Pick lots that survive a bad run, test on your own broker, and only risk money you can afford to lose. No DLLs or extra files are needed.
If you have run it, an honest review of how it behaved for you, good or bad, helps the next trader decide.
