Chris Amow
8ca624d435
The confluence engine had nothing to work with. Daily moving averages were the only level source, and they sat 163 to 697 points from price, so every cluster had exactly one member and no alert could ever fire. Two new sources, chosen for having a real following — the engine is a bet that many participants watch the same price, which is what makes a level hold: - Prior day high/low/close, from the last *closed* daily bar so mid-session the levels do not silently switch to today's own developing range. Full daily weight rather than the 0.75 average discount: a traded high is structure, not a derived average. - Session VWAP, anchored to the 18:00 ET open like the daily bars. Institutional execution is benchmarked against it, and zero-volume overnight minutes are skipped rather than dividing by zero. Both are stamped 1d, so they get their own colours to stay distinguishable from the daily averages. Prior-day levels draw as price lines, which span the chart and label the axis instead of relying on bar-index interpolation. VWAP re-prices every minute while a daily average carries hundreds of points and changes once a session, so broadcasting the whole level set on the VWAP cadence would have pushed the entire history every minute. Levels now go out as a delta that clients merge by id. Adding the levels then exposed two defects that had been invisible while nothing could cluster: - Cluster identity was sha1(side + round(center / tolerance)), and tolerance derives from ATR, so it changed every bar. The same zone was continually issued a new id, never matched the cooldown table, and the cooldown did nothing. Identity is now the set of converging levels. - Alert suppression keyed on that identity, so a level drifting in or out of a group read as a new zone. It now suppresses by proximity: two zones within an ATR are the same zone, and the strongest is the one reported. Over six replayed sessions at threshold 28 that is 247 alerts, then 54, then 40; raising the cooldown to 4h — which only affects repeats of the same area, never a genuinely new zone — gives 17 total with a worst session of 9. calibrate_alerts.py now sweeps threshold and cooldown together in one pass, since the threshold turns out to be quantised and nearly useless as a control. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
112 lines
4.9 KiB
Python
112 lines
4.9 KiB
Python
"""Replay Yahoo's available minute tape and report alerts per CME session.
|
|
|
|
Sweeps a range of thresholds in a single pass rather than testing only the
|
|
configured one: the useful question is where the alert rate crosses from silent
|
|
to noisy, which a single number cannot show.
|
|
|
|
Manual trendlines are deliberately excluded — they are user data, and a
|
|
threshold calibrated against one person's drawings would not transfer.
|
|
"""
|
|
import asyncio
|
|
from collections import Counter
|
|
|
|
from app.analysis.alerts import AlertEngine
|
|
from app.analysis.confluence import cluster_levels
|
|
from app.analysis.horizontals import build_prior_day_levels
|
|
from app.analysis.indicators import atr
|
|
from app.analysis.moving_averages import build_ma_levels
|
|
from app.analysis.vwap import build_vwap_level
|
|
from app.bars.aggregator import Aggregator
|
|
from app.bars.models import Timeframe
|
|
from app.bars.session import bucket_start
|
|
from app.bars.store import InMemoryBarStore
|
|
from app.config import Settings
|
|
from app.market.yahoo import YahooSource
|
|
|
|
THRESHOLDS = (12, 16, 20, 24, 28, 32, 40)
|
|
# Scores are sums of 12s and 16s, so the threshold is quantised and blunt:
|
|
# several values behave identically and then it falls to zero. Cooldown is the
|
|
# finer control over how often a zone price is chopping around repeats itself.
|
|
COOLDOWNS = (900, 1800, 3600, 7200, 14400)
|
|
|
|
|
|
async def main() -> None:
|
|
settings = Settings()
|
|
source = YahooSource(settings.yahoo_poll_seconds)
|
|
hourly, minutes = await asyncio.gather(
|
|
source.history(settings.yahoo_symbol, Timeframe.H1, range_=settings.seed_1h_range),
|
|
source.history(settings.yahoo_symbol, Timeframe.M1, range_=settings.seed_1m_range),
|
|
)
|
|
if not minutes:
|
|
raise RuntimeError("Yahoo returned no minute tape")
|
|
|
|
aggregator = Aggregator(settings.enabled_timeframes)
|
|
store = InMemoryBarStore(25_000)
|
|
# Keyed on (threshold, cooldown) so one replay pass measures both sweeps.
|
|
combos = [(threshold, settings.alert_cooldown_seconds) for threshold in THRESHOLDS]
|
|
combos += [
|
|
(settings.confluence_min_score, cooldown)
|
|
for cooldown in COOLDOWNS
|
|
if cooldown != settings.alert_cooldown_seconds
|
|
]
|
|
engines = {combo: AlertEngine(combo[0], combo[1]) for combo in combos}
|
|
counts: dict[tuple, Counter[int]] = {combo: Counter() for combo in combos}
|
|
|
|
ma_levels: list = []
|
|
cutoff = minutes[0].t
|
|
for source_bar in [bar for bar in hourly if bar.t < cutoff] + minutes:
|
|
for bar in aggregator.update(source_bar):
|
|
store.put(bar)
|
|
if settings.ma_sets.get(bar.tf):
|
|
ma_levels = build_ma_levels(
|
|
{tf: store.get(tf) for tf in settings.ma_sets}, settings.ma_sets
|
|
)
|
|
if bar.tf is not Timeframe.M1 or not bar.closed:
|
|
continue
|
|
atr_values = atr(store.get(Timeframe.M15), 14)
|
|
atr15 = next((value for value in reversed(atr_values) if value is not None), 0.0)
|
|
levels = (
|
|
ma_levels
|
|
+ build_prior_day_levels(store.get(Timeframe.D1), bar.c)
|
|
+ build_vwap_level(store.get(Timeframe.M1))
|
|
)
|
|
# Clustering is threshold-independent, so it is done once and the
|
|
# result fed to every engine.
|
|
clusters = cluster_levels(levels, bar.t, bar.c, atr15)
|
|
session = bucket_start(bar.t, Timeframe.D1)
|
|
for combo, engine in engines.items():
|
|
alerts = engine.evaluate(clusters, bar.c, atr15, bar.t, settings.yahoo_symbol)
|
|
counts[combo][session] += len(alerts)
|
|
|
|
sessions = sorted({session for counter in counts.values() for session in counter})
|
|
print(f"minute_bars={len(minutes)} sessions={len(sessions)}")
|
|
|
|
def report(title: str, selected: list[tuple]) -> None:
|
|
print(f"\n{title}")
|
|
print(f"{'threshold':>9} {'cooldown':>9} {'total':>6} {'max/sess':>9} per-session")
|
|
for combo in selected:
|
|
per_session = [counts[combo][session] for session in sessions]
|
|
configured = combo == (settings.confluence_min_score, settings.alert_cooldown_seconds)
|
|
print(
|
|
f"{combo[0]:>9g} {combo[1]:>9} {sum(per_session):>6} "
|
|
f"{max(per_session, default=0):>9} "
|
|
f"{', '.join(str(value) for value in per_session)}"
|
|
f"{' <- configured' if configured else ''}"
|
|
)
|
|
|
|
report(
|
|
f"threshold sweep (cooldown={settings.alert_cooldown_seconds}s)",
|
|
[(threshold, settings.alert_cooldown_seconds) for threshold in THRESHOLDS],
|
|
)
|
|
report(
|
|
f"cooldown sweep (threshold={settings.confluence_min_score:g})",
|
|
sorted(
|
|
{(settings.confluence_min_score, cooldown) for cooldown in COOLDOWNS}
|
|
| {(settings.confluence_min_score, settings.alert_cooldown_seconds)},
|
|
key=lambda combo: combo[1],
|
|
),
|
|
)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
asyncio.run(main())
|