ayantrader
832e0af648
Size the scaling-law control to EUR/USD tick volatility, run the random-walk study over 200 seeds, drop the tester and terminal-close hooks, rename the reference list to imbalanced, and update the README with the findings.
77 Zeilen
4,4 KiB
Markdown
77 Zeilen
4,4 KiB
Markdown
# IntrinsicTime
|
|
|
|
The directional-change operator, the intrinsic-time scaling laws, and the Alpha
|
|
Engine coastline trader, implemented in MQL5.
|
|
|
|
Companion code for the MQL5 article: https://www.mql5.com/en/articles/23814
|
|
|
|
## What it does
|
|
|
|
Physical time is an arbitrary clock for a market. Intrinsic time replaces it:
|
|
the clock only ticks when price reverses by a fixed threshold, so quiet periods
|
|
compress and active ones stretch. `DcOperator.mqh` is that clock: an online
|
|
operator that dissects a tick stream into directional-change and overshoot
|
|
legs, one price at a time, with no look-ahead.
|
|
|
|
That reframing comes with published scaling laws, which are empirical
|
|
regularities relating threshold size to event counts and move lengths.
|
|
`DcOS_ScalingLaws.mq5` measures them on your own broker's ticks and fits the
|
|
same laws on a Gaussian random walk as a control. On 17.8 million EUR/USD ticks
|
|
the market's exponents match the paper's closely (count law -1.944 against the
|
|
paper's -1.908, coastline -0.978 against -0.940). The control is the important
|
|
part. A walk sized to the market, one step per price change with the market's
|
|
own step volatility, reproduces the same exponents (-1.974 and -0.969). At these
|
|
scales the laws hold for pure noise too, so they do not by themselves reveal
|
|
market structure. A walk built with the paper's much coarser fixed step gives
|
|
shallower slopes and a false-looking gap, which is why the script's defaults
|
|
size the walk to EUR/USD.
|
|
|
|
The Alpha Engine is the trading side: a counter-trend coastline trader that adds
|
|
to a position at each adverse intrinsic event and trims each add at a profit of
|
|
delta, with inventory-skewed thresholds and a liquidity indicator around it.
|
|
`Ae_RandomWalk.mq5` runs the reference engine on 200 seeded random walks. Its
|
|
realised profit is positive on all 200, but its total profit, with the open
|
|
cascade marked at the last price, averages zero (+0.22, standard error 0.36):
|
|
the method turns noise into many small closed wins and a few large open losses,
|
|
not into an edge.
|
|
|
|
`AlphaEngine.mq5` is the Expert Advisor, eight limit-order agents on a hedging
|
|
account. On EUR/USD from 2 February to 31 July 2026 at the default inputs it
|
|
returned 10.68% with a 10.10% equity drawdown. That is +1,934 from take-profits,
|
|
-824 when the tester closed 24 leftover positions at the end of the run, and
|
|
-42 commission, with swap (-620 in total) already inside the first two figures.
|
|
The live agent has no whole-position exit, so a cascade that never recovers is
|
|
only halted, not closed.
|
|
|
|
## Layout
|
|
|
|
```
|
|
Include/IntrinsicTime/DcTypes.mqh directional-change enums and event record
|
|
Include/IntrinsicTime/DcOperator.mqh CDcOS, the directional-change/overshoot operator
|
|
Include/IntrinsicTime/ScalingLaws.mqh operator bank, log-log fit, random-walk control
|
|
Include/IntrinsicTime/AeTypes.mqh tick structure and intrinsic-event codes
|
|
Include/IntrinsicTime/AeRunner.mqh CAeRunner, the log-threshold event runner
|
|
Include/IntrinsicTime/AeLiquidity.mqh CAeLocalLiquidity, the liquidity indicator L
|
|
Include/IntrinsicTime/AeLimitOrder.mqh reference limit order with de-cascade accounting
|
|
Include/IntrinsicTime/AeCoastlineTrader.mqh the offline reference coastline trader
|
|
Include/IntrinsicTime/AeAlphaEngine.mqh the eight-agent ensemble, offline reference form
|
|
Include/IntrinsicTime/AeLiveTrader.mqh CAeLiveTrader, one live agent on real limit orders
|
|
Experts/IntrinsicTime/AlphaEngine.mq5 the Expert Advisor (hedging account required)
|
|
Scripts/IntrinsicTime/DcOS_SelfTest.mq5 known-answer tests for the operator
|
|
Scripts/IntrinsicTime/DcOS_ScalingLaws.mq5 scaling laws on live ticks against the control
|
|
Scripts/IntrinsicTime/Ae_SelfTest.mq5 known-answer tests for the trading model
|
|
Scripts/IntrinsicTime/Ae_RandomWalk.mq5 realised and total profit on random walks
|
|
```
|
|
|
|
Run the two self-tests first, then the scaling-law script on a fully synced
|
|
EUR/USD chart. For another symbol, set the walk's length, starting level and
|
|
per-step sigma to that symbol's own tick data before comparing exponents.
|
|
|
|
## Disclaimer
|
|
|
|
Educational code. The random-walk results show that the engine's closed-trade
|
|
profit is not evidence of an edge, and the backtest covers one pair over one
|
|
six-month window, so it demonstrates a working program rather than a strategy
|
|
with proven expectancy. Past behaviour of any model or dataset says nothing
|
|
about future results. Test on your own data and broker conditions before
|
|
drawing conclusions.
|