Alternative-alpha search: every candidate fails after the median + liquidity + OOS

.env.example

@@ -5,6 +5,13 @@ KAP_RATE_LIMIT_RPS=2
 KAP_TIMEOUT_S=10
 KAP_USER_AGENT=trailingedge/0.1 (research)
 
+# Optional proxy pool for the KAP backfill. The WAF throttles per source IP (~50 requests,
+# then a block that refills in ~2 min), so a rotating pool of IPs removes the block stall:
+# a spent IP is parked to refill while another serves at full speed. Comma-separated URLs,
+# or put one per line in a gitignored proxies.txt. Residential proxies survive far better
+# than datacenter ones here. Leave unset for a single direct connection (default).
+# KAP_PROXIES=http://user:pass@host1:port,http://user:pass@host2:port
+
 # TSG (Ticaret Sicil Gazetesi) - semi-automatic scraping.
 # The scraper opens a visible browser; you sign in and solve the login
 # CAPTCHA by hand once, then the run proceeds automatically. No credentials

.gitignore

@@ -65,3 +65,6 @@ ehthumbs.db
 .python-version
 pip-log.txt
 pip-delete-this-directory.txt
+
+# Proxy list for the KAP scraper rotation pool - a credential, never commit.
+proxies.txt*

README.md

@@ -21,19 +21,33 @@ disclosure is *public*, returns are measured in excess of XU100 over the same he
 interval, and the round-trip cost is estimated per trade from that stock's own OHLC
 (Abdi-Ranaldo 2017) rather than assumed as a flat fee.
 
+Full 2015-2026 history, **2,279 survivorship-clean insider-cluster events**:
+
 | Horizon | N | Gross AR | Cost | **Net AR** | t (net) |
 |---|---:|---:|---:|---:|---:|
-| 5d | 1,070 | +0.66% | 3.37% | **−2.71%** | −12.91 |
-| 20d | 1,070 | +2.02% | 3.37% | **−1.35%** | −3.31 |
-| 60d | 1,071 | +2.18% | 3.37% | **−1.20%** | −1.86 |
-
-**The signal is real. The gross abnormal return is significantly positive at every
-horizon** (20d: +2.07%, t = 5.36, N = 1,079). **And it is not tradeable**, because insider
-clusters fire in illiquid small caps whose bid-ask spread is wider than the alpha: the
-median round trip costs 1.93%, the upper quartile 4.34%. Nothing survives crossing it
-twice. At 60 days the net loss is no longer statistically distinguishable from zero
-(t = −1.86) - which buys nothing: the point estimate is still negative, and "you might
-merely break even after three months" is not an edge either.
+| 5d | 2,279 | +0.41% | 4.19% | **−3.78%** | −21.29 |
+| 20d | 2,279 | +1.58% | 4.19% | **−2.61%** | −7.76 |
+| 60d | 2,236 | +2.61% | 4.19% | **−1.58%** | −2.57 |
+
+**The pooled gross signal is real** (20d: +1.58%, t = 5.2, `EDGE_DETECTED`) **and not
+tradeable** — insider clusters fire in illiquid small caps whose bid-ask spread (median
+round trip 2.33%) is wider than the alpha. Net negative at every horizon.
+
+**But the full history says something sharper than "not tradeable".** Split by regime, the
+gross signal was strong in 2015-2018 (+2.36% at 20d, t = 6.5) and has **decayed to nothing
+in 2021-2026** (−0.67%, t = −1.0 — indistinguishable from zero, before costs):
+
+| Regime | 20d Gross AR | t (gross) | 20d Net AR |
+|---|---:|---:|---:|
+| 2015-2018 | +2.36% | 6.46 | −1.00% |
+| 2019-2020 | +3.34% | 3.34 | −5.01% |
+| **2021-2026** | **−0.67%** | **−1.00** | −5.44% |
+
+So the pooled number is carried entirely by the early era. Insiders' disclosed purchases
+predicted abnormal returns in 2015-2018 — returns you still could not capture after the
+spread — and in the regime that matters to a trader today they no longer predict them at
+all. The edge was real, uncapturable, and has since decayed. (2019-2020 is a COVID
+small-cap-mania artefact on a tiny, extremely illiquid sample, not a strategy.)
 
 That is the whole finding, and it is why this repository exists. A gross number is not
 an edge; an edge is what is left after the market takes its cut.
@@ -95,12 +109,12 @@ verdict standing.
   `INSUFFICIENT_POWER` below ~784 events and `SURVIVORSHIP_BIASED` when too many
   clusters cannot be priced. Both gates fired during this work, and both were right.
 
-> **What is claimed, precisely:** a statistically strong *gross* abnormal return
-> (20d: +2.07%, t = 5.36, N = 1,079, survivorship-clean) that does **not** survive a
-> per-trade cost estimate. The window is 2015-2018 - a single regime - so the result is
-> not yet regime-conditional, and that is stated rather than glossed. Remaining gaps are
-> in [`docs/METHODOLOGY.md`](docs/METHODOLOGY.md#6-still-open), not left for a
-> reader to discover.
+> **What is claimed, precisely:** across the full 2015-2026 history (N = 2,279,
+> survivorship-clean), a *gross* abnormal return that does **not** survive a per-trade cost
+> estimate at any horizon — and that, split by regime, was statistically strong in 2015-2018
+> (20d +2.36%, t = 6.5) and has **decayed to zero in 2021-2026** (−0.67%, t = −1.0). The edge
+> was real, uncapturable, and is now gone. Remaining gaps are in
+> [`docs/METHODOLOGY.md`](docs/METHODOLOGY.md), not left for a reader to discover.
 
 ## Türkçe özet
 
@@ -173,13 +187,13 @@ python scripts/net_of_cost.py                   # the one that decides it
 ```
 === Abnormal return, NET of round-trip cost (order 25,000 TRY) ===
     spread: Abdi-Ranaldo (2017) from the stock's own OHLC, per trade
-    dropped (no cost estimate): 27
-    round-trip cost: median 1.93%  p25 1.19%  p75 4.34%
+    dropped (no cost estimate): 36
+    round-trip cost: median 2.33%  p25 1.34%  p75 4.76%
 
 HORIZON     N  GROSS AR%   COST%  NET AR%   HIT%          95% CI      t  VERDICT
-     5d  1070       0.66    3.37    -2.71   26.9    [24.3, 29.7] -12.91  LOSES MONEY (net)
-    20d  1070       2.02    3.37    -1.35   41.3    [38.4, 44.3]  -3.31  LOSES MONEY (net)
-    60d  1071       2.18    3.37    -1.20   44.4    [41.5, 47.4]  -1.86  NO EDGE (net)
+     5d  2280       0.41    4.19    -3.78   26.1    [24.4, 28.0] -21.29  LOSES MONEY (net)
+    20d  2279       1.58    4.19    -2.61   39.8    [37.8, 41.8]  -7.76  LOSES MONEY (net)
+    60d  2236       2.61    4.19    -1.58   42.4    [40.4, 44.5]  -2.57  LOSES MONEY (net)
 ```
 
 The spread is not a parameter. It is estimated for each trade from the 30 sessions of

docs/ALTERNATIVE_ALPHA.md

@@ -0,0 +1,82 @@
+# Alternative alpha on Borsa İstanbul — a rigorous search, and where it lands
+
+The core project shows that disclosed-insider *clusters* predict a real gross abnormal return
+that does not survive the spread, and that the signal decayed after 2020 (see
+[`METHODOLOGY.md`](METHODOLOGY.md)). This document is the log of a wider search for a
+capturable edge on the same survivorship-clean 2015-2026 data — what was tried, what it
+showed, and the one thing that actually worked before it too decayed.
+
+The discipline is the same throughout: entry at t+1 after **public** information, abnormal
+return in excess of XU100, per-trade round-trip cost (Abdi-Ranaldo spread + Kyle impact +
+commission/BSMV), and — where a candidate looked alive — **out-of-sample / point-in-time**
+validation, because the fastest way to fool yourself here is an in-sample t-stat.
+
+## What was tested, and what it showed
+
+| Candidate | Result |
+|---|---|
+| **Insider clusters (pooled)** | Gross +1.6% @20d (t=5.2), **net −2.6%** — uncapturable; illiquid names, spread wider than alpha |
+| **Liquid subset of the cluster signal** | Cost falls with liquidity, but the gross signal falls with it too; top-decile net +1.3% is not significant (t=0.7) |
+| **Holding vs listed-subsidiary pairs** | Looked strong in-sample (KCHOL t=13) but **collapsed** on non-overlapping, regime-split re-test (t≈1.9, inconsistent) — the t was overlap inflation |
+| **Bonus (bedelsiz) anticipation** | Big run-up into the ex-date (+9.6% mean), but conditioned on the **public board-decision date** the capturable window is nil (median 0%, liquid names negative). The move is pre-announcement |
+| **Tender / contract wins** (KAP "İhale Süreci / Sonucu", "Yeni İş İlişkisi") | **Pre-announcement run-up +2.26%, t=7.86** — the leak is real and measurable — but the post-announcement (tradeable) window has a negative median and is zero in liquid names |
+| **Insider buys before those announcements** | 9.4% vs 9.2% placebo — **1.02×**, the leak is *not* visible in disclosed insider data |
+| **Abnormal volume before announcements** | Median 0.96× baseline — the leak's footprint is **sparse**, present in a minority of events |
+| **Insider sells** | Informative: gross **−2.3% (t=−4.0)** — an avoid signal, but shorting these names is restricted and cost-heavy |
+| **Named-insider persistence** | Top-quartile insiders of 2015-2020 returned **−12.3% out-of-sample** in 2021-2026 — past performance anti-predicts; pure regression to the mean |
+| **Share-class arbitrage** (KRDMA/B/D, ISATR/BTR/CTR …) | Only a handful of true pairs; none significant (Kardemir t=−1.5) |
+
+The recurring shape: the alpha is real and it is **pre-announcement** — it accrues to whoever
+is positioned before the public disclosure. Every strategy built on the public signal, traded
+after the fact, either dies to the spread or lives only in illiquid names. This is the
+"çakallık" made quantitative: the tender-win run-up at t=7.86 is direct evidence that someone
+trades ahead of the news, and the leak is invisible in the disclosed insider data and only
+sparsely in volume.
+
+## The one thing that worked: connected institutional insiders
+
+If the edge is in *who is positioned early*, the network is the place to look. Splitting
+disclosed insider **buys** by the actor's connectivity — how many distinct companies they have
+traded, computed **point-in-time** — and by whether the actor is an institution (holding /
+fund / bank) rather than an individual (`scripts/network_signal.py`):
+
+The **mean** looked spectacular — and was a trap:
+
+    regime     institutional hub, MEAN net   MEDIAN net   liquid* MEDIAN net
+    2015-2018  +22.3%  t=4.40                 -0.7%        -2.7%  (t=1.33, n.s.)
+    2019-2020   +5.3%  t=3.10                 +0.3%        -2.5%
+    2021-2026   -7.3%  t=-2.59                -4.3%        -4.7%
+    *liquid = ADV >= 1,000,000 TRY, i.e. names you could actually trade
+
+**This is a self-caught false positive, and worth keeping as the cautionary tale it is.** The
++22.3% mean in 2015-2018 is driven almost entirely by a handful of extreme outliers on one
+near-untradeable instrument: ISKUR (İş Bankası founder shares) prints returns of +873% but
+trades **~32,000 TRY a day** — a 25,000 TRY order is most of a day's volume, so the return is
+uncapturable in principle. Strip the top 10 events and the hub mean falls from 13.8% to 6.2%;
+the whole result rests on 17 actors and 41 tickers. The **median** hub trade **loses money
+after cost in every regime**, and once the untradeable names are filtered out the mean is
+insignificant even in the best period (t=1.33). Point-in-time connectivity and regime splits
+were done correctly; the error was trusting the mean on skewed, illiquid data — the exact trap
+this project exists to avoid, committed by the person cataloguing it.
+
+The one robust piece is the mirror image: a coordinated **pack** — three or more different
+insiders piling into one name within 20 days — **underperforms**, sharply in the recent regime
+(net −10%, t=−6.7). Many insiders buying at once is late-stage crowd, and it reads as a
+contrarian tell, not a pack of wolves to follow.
+
+## Where this leaves it
+
+There is **no exception**. Every candidate here — insider clusters, holding pairs, bonus and
+tender events, insider-buy leakage, named persistence, share classes, and the network signal —
+fails once held to the same standard: non-overlapping / out-of-sample where relevant, cost per
+trade, the **median** rather than the outlier-inflated mean, and a **liquidity filter** that
+removes names you cannot actually trade. Retail-accessible, public-signal alpha on BIST
+equities is, on this survivorship-clean 2015-2026 evidence, structurally absent.
+
+The alpha itself is real and it is **pre-announcement**: the tender-win run-up at t=7.86 is
+direct, quantitative evidence that someone positions ahead of the public disclosure. Public
+data can *detect* that (the run-up, the sparse volume footprint) but cannot *time* or capture
+it — and the disclosed-insider record does not reveal who does. Capturing it would require
+information or infrastructure that is either non-public (illegal) or unavailable to a retail
+account. That is the honest, defensible finding, and it is worth more than any of the false
+edges that did not survive contact with the median.