Working notes
from the desk.
Methodology, backtests, and risk philosophy from the Shishin quant team. New material every few days. No paywall, no email gate, no preface about why you should subscribe. Just the work.
Mean reversion vs momentum the two forces every strategy bets on
Momentum says trends persist and mean reversion says extremes snap back. Both are real: they just work on different time horizons. Here is how they fit.
How to vet a track record.
Before trusting a signal service's track record: is it survivorship-free, point-in-time, net of costs, significant, and does it show the losers?
Compounding and volatility drag: why arithmetic and geometric returns differ why arithmetic and geometric returns differ
Arithmetic and geometric returns differ because wealth compounds. Volatility drag is roughly half the variance, and losses cost more than equal gains.
The episodic pivot. The gap is the announcement.
The episodic pivot (EP) explained: a stock gapping out of a quiet base on an earnings surprise, and why earnings-confirmed gaps run while technical gaps fade.
Risk of ruin: the math of how position size kills accounts why a positive-expectancy strategy can still go broke, and why surviving comes before compounding
Risk of ruin is the chance a losing streak drains an account past recovery before its edge compounds. Why position size, not win rate, dominates it.
Inside Byakko. The defensive engine, by the numbers.
A deep look at Byakko, Shishin's defensive engine: the only one that wins over half its trades, and how it earns through downtapes.
The Kelly criterion and fractional Kelly
What the Kelly criterion is, how edge-over-odds bet sizing works, why full Kelly is too aggressive, and why practitioners run fractional Kelly.
The Cup-and-Handle Pattern, and How to Read Its Depth
How the cup-and-handle base is defined and read: the rounded cup, the shallow upper-half handle, cup depth, and the tells of a healthy versus faulty base.
What is volatility? The most quoted, most misread number in markets.
Volatility is the dispersion of an asset's returns, measured as their standard deviation and annualized by root-time. What it does and does not tell you.
Inside Seiryū. The recovery engine, by the numbers.
A deep look at Seiryū, Shishin's bear-market recovery engine: why it trades just 24 times in five years yet earns the most per day on duty.
How accurate are quant stock signals?
Accuracy is the wrong question: a good systematic strategy can be right under half the time and still profit. Why expectancy beats hit rate.
How a base tightens before it breaks.
The volatility contraction pattern (VCP), Mark Minervini's tightening base: progressively shallower pullbacks, drying volume, and a coil toward the pivot.
When momentum crashes.
Momentum's darkest episodes are state-dependent and partly forecastable. What Daniel and Moskowitz found, and how a regime gate changes the risk.
Dollar volume. The liquidity gate.
Dollar volume (price times shares) is the liquidity gate: why a signal in a thin stock is untradeable, and how it sets your size and slippage ceilings.
We don't trade the open. We trade the close.
The open is the day's most gap-prone print. Why the live system fills entries in the closing auction instead, and how the backtest stays conservative.
What is overfitting in trading? Brilliant on history, broken live.
Overfitting is tuning a strategy to noise in its backtest, so it looks brilliant on history and fails live. How it sneaks in, and how to catch it.
The flag breakout: reading the continuation setup
The flag continuation pattern explained: the pole, the tight low-volume pause, the volume signature, and how a healthy flag differs from one that fails.
What is ATR (Average True Range)? A volatility measure, not a direction.
Average True Range (ATR) measures how much a stock typically moves, gaps included. How it's calculated, what it's for, and how Shishin sizes stops with it.
Why most backtests lie. And how to read past it.
The mechanical ways a backtest inflates the past, look-ahead, overfitting, the in-sample trap, and why walk-forward and live trading are the honest tests.
What is the Sortino ratio? The downside-only Sharpe.
Sortino judges return against downside deviation only, not total volatility: the fairer lens for a right-skewed momentum strategy. Plus the Calmar ratio.
Where a stock is in its own cycle.
The same signal can mean opposite things: a fresh 50-day-MA cross differs from a year above it. How Shishin's setup-state classifier captures the difference.
What a pre-market scan actually computes.
What a real pre-market scan computes, trend, volatility, momentum, structure, liquidity, and how dozens of measurements become one daily ranked list.
Volatility-aware stops, vs fixed percentages.
The stop-placement sweep, with receipts: ATR-derived stops with an upper cap beat every fixed-percentage variant we tested. The principle, not the parameters.
Quant vs discretionary, and where each wins.
Rules vs judgment: where discretionary genuinely wins, where systematic does, and why the real divide is consistency versus adaptability.
Four guardians, one mental model.
Why Shishin's four engines carry the names of the Four Symbols of East Asian cosmology: one word for the entire market posture the system is taking today.
Why one strategy can't work in every market.
Markets move through states; a strategy tuned for one fails in another. The case for matching approach to regime, without whipsawing on the turns.
Shishin vs Danelfin, attested, not advertised.
Shishin vs Danelfin, a fair non-advisory comparison. The wedge is proof: an attested forward record versus backtested, self-reported headline numbers.
Shishin vs Motley Fool, auditable, not just advertised.
Shishin vs Motley Fool Stock Advisor: a buy-and-hold recommendation newsletter versus a free, non-advisory ranked board with an externally attested record.
Shishin vs Zacks, verifiable, not just ranked.
Shishin vs Zacks: a fair comparison. The Zacks Rank is a respected earnings-revision signal with self-reported returns; Shishin's record is attested.
Shishin vs TipRanks, independent, and checkable.
Shishin vs TipRanks: a fair comparison. Smart Score aggregates lagging analyst consensus; Shishin is a price-based signal on an open, attested board.
Shishin vs Benzinga, news, or a tracked signal.
Shishin vs Benzinga: a fair comparison. Benzinga Pro is a news terminal, not a signal service; Shishin is a rules-based board with an attested record.
Shishin vs IBD, free, open, and attested.
Shishin vs IBD: a fair comparison. CAN SLIM is a breakout institution, but paywalled with self-reported returns; Shishin's board is free and attested.
Shishin vs Tickeron, one record you can check.
Shishin vs Tickeron: a fair comparison. Tickeron sells many AI bots with no single forward-tested record; Shishin publishes one live, attested record.
Shishin vs Trade Ideas, attested, not simulated.
Shishin vs Trade Ideas: both research, not advice. The wedge is proof and price: an attested every-call record versus Holly's simulated 'biggest hits'.
Is Motley Fool legit? yes, but worth it is a different question.
Is Motley Fool legit? Yes, a real publisher since 2002, not a scam. Whether Stock Advisor is worth it depends on your holding behavior.
Is Danelfin legit? real tool, backtested proof.
Yes, Danelfin is a legit non-advisory AI scorer with real explainability and a free tier. The catch: its headline numbers are largely backtested.
Is Tickeron worth it? the backtest-versus-live gap.
A standalone review of Tickeron's AI trading bots: real strengths, the backtest-versus-live gap, and what to check before funding one.
Is Trade Ideas worth it? a standalone verdict on Holly AI.
A plain, non-advisory verdict on Trade Ideas and Holly AI: the genuine strengths, the simulated-results catch, and who the premium is actually worth it for.
A track record you can verify, not trust.
Most stock signals are self-reported, not verifiable. What an attested, tamper-evident track record is, why the field avoids it, and who has one.
Best AI stock pickers, ranked by what you can verify.
A non-advisory ranking of the main AI stock pickers by one axis: can you independently verify the record? Backtested vs simulated vs live vs attested.
The most transparent stock-signal services, scored on what you can verify.
Transparency scored as four checkable properties: a full winners-and-losers record, external attestation, a legible method, and non-advisory framing.
Free stock signals, and what each tier really hands you.
The big stock-pick brands paywall their picks. A fair, non-advisory look at genuinely free signal tiers, plus a full board an AI assistant can actually read.
What is alpha (vs beta)? Skill, or just the market?
Beta is the return from cheap market exposure; alpha is the skill-based excess that survives a regression against a benchmark. How to tell the two apart.
Inside Genbu. The quality engine, by the numbers.
A deep look at Genbu, Shishin's small-cap quality engine: its regime gate, how it picks and exits, and why a 35.7% win rate still made money over 42 trades.
The boring middle. Where most of the work happens.
Most trades are neither big winners nor blow-ups; they cluster near zero. Why the discipline of a momentum system lives in how it treats that boring middle.
What is RSI? The oscillator, not relative strength.
Wilder's 0-100 momentum oscillator: what overbought and oversold really mean, why RSI is not 'relative strength', and why it's context, not a buy trigger.
The names that aren’t there, and what we do about them.
Why most equity backtests are contaminated by survivor selection, and the five countermeasures that keep Shishin's universe honest at every historical date.
Inside Suzaku. The breakout engine, by the numbers.
A deep look at Suzaku, Shishin's breakout engine: the regime gate, how it picks and exits trades, and what its 131 backtested trades reveal.
What is market breadth? What the index level hides.
Market breadth: how many stocks are actually participating in a move, not just a few mega-caps carrying the index. What it reveals, and its limits.
The seed loss. Why small drift is catastrophic.
How a small early error compounded over five years into a half-million-dollar gap, and why operating discipline matters more than any single rule.
The best stock signal services, honestly tested.
A buyer's guide to stock signal services, the seven tests that separate a real edge from a good-looking backtest, plus the field compared by approach.
Do stock signals actually work?
Some stock signals work, most don't, and the difference is expectancy net of costs, reproducibility, and whether you actually follow them. How to tell.
Position sizing, by conviction.
How a composite score becomes a position size: top-up logic, score-conviction mapping, and why the highest-conviction names get the most capital.
Regime-adaptive, or just relabelled?
What makes a stock-signal strategy regime-adaptive, why static strategies fail across regimes, and five questions to ask when evaluating one.
What is momentum investing? The anomaly that shouldn't work, and does.
What momentum investing is, why the winners-keep-winning anomaly persists across decades and asset classes, and the regime-turn crash that can undo it.
Paper-trade in public.
Why the Shishin live bot publishes every NAV tick, open position, and fill. The compounding edge of forced honesty in systematic trading.
Rebalancing, and adding to winners.
Trimming to fund better ideas and gated top-ups into winners: what Shishin tested in capital rotation, why naive add-to-winners fails, and why it stayed out.
The macro classifier. Seven regimes, one decision.
Inside the breadth-driven regime classifier: seven regimes, daily transitions, and how universe-wide MA stack readings drive capital allocation.
The idle-cash problem, and a momentum fix.
Idle cash costs returns. How Shishin parks it in a momentum-ranked ETF rotation sleeve that improved return, Sharpe, and drawdown together in backtest.
What a Sharpe of 1, 2, or 3 actually means.
What the Sharpe ratio measures, what a 1 vs 2 vs 3 actually tells you, and the trade-off nobody mentions: a higher Sharpe usually means a lower return.
How breakout setups work, and why most fail.
The anatomy of a breakout, base, pivot, expansion, the mechanic behind the ones that work, and the structural reasons most breakouts fail.
Composite scoring. Eight features, one number.
How Shishin's composite is built: eight weighted sub-scores across position, volatility, breakout, liquidity and sector. Deterministic and transparent.
Return alone, Sharpe alone, drawdown alone. None of them.
No single metric is enough: return, Sharpe, and drawdown each fail when optimised alone. A sound strategy improves all three together.
How a stock signal is made, and what it is not.
A trading signal is ranked, rule-based research, not a buy command. How quant signals are scanned, scored, and ranked, and how to spot a real one.
Drawdowns are a feature, not a side-effect.
A strategy's return tells only half the story. See what drawdown measures, and why we optimize for the shape of the equity curve, not just the endpoint.
Four engines for four regimes. One discipline.
Inside the four-engine architecture: Genbu, Suzaku, Byakko, and Seiryū, each tuned for a different market regime, gated by a breadth-driven macro classifier.