thetrading.tools
RiskUpdated daily after the close · model v1.0, frozen 2026-09-18

Risk On / Risk Off: A Daily SPY Long/Short Signal

One fixed rule, three inputs, one target a day. It combines the 63-session advance share, the velocity of ES and NQ session control and the 10-session change in the VIX. Risk On targets a long SPY position and Risk Off targets a short one, filled at the next open. The simulated record below is in-sample and this page keeps saying so.

Latest reading

As of the September 18, 2026 close, the model reads Risk On: it targets a long SPY position at the next session's open. That signal has stood since September 15, 2026, 4 sessions ago.

Sources, methodology & freshnessLast updated 2026-09-18 · Open ↓
Source
SPY daily OHLC, the published 63-session advance share, ES/NQ session candle sums and the VIX close — all from this site’s own daily data
Methodology
Three bounded inputs averaged with equal weight; long above −0.15, short below −0.15, previous signal retained on a tie or a missing input. Signals fill at the next SPY open with 2bp per fill and 0.5% annual short borrow. Version 1 was selected on full-period return through September 17, 2026, so the headline record is in-sample
Updates
Daily, after the closeLast: 2026-09-18
Maintained & reviewed by Yuriy Matso — methodology shown on the page.
Next-open target2026-09-18 · close
Risk On
score 55.7

Targets a long SPY position at the next session's open. Held since 2026-09-15 (4 sessions). The 0–100 score crosses at 42.5; it is not a probability.

SPY close
$761.69
Position held
Long
Completed trades
275

Risk Off means the rule targets a short. It is not a judgment that markets are unsafe, and the signal can be wrong for as long as it is held.

01

What the three inputs say today

Each input is clipped to a value between −100 and +100 and the three are averaged with equal weight. A positive contribution pushes toward Risk On. At publication (2026-09-18) the average sat at 55.7 on the 0–100 scale, against a decision boundary of 42.5.

02

Simulated account since 2019

2019-01-02 → 2026-09-18 · 1939 sessions · one shared range and crosshair

Account equity

2026-09-18Model $87,095SPY $30,954
Account equity: Risk On / Risk Off compared with holding SPY$0k$20k$40k$60k$80k$100k2020202220242026
$10,000 invested at the first session’s open. Net of modeled costs; dividends excluded. Green: model. Gray: holding SPY.

SPY with the position held

2026-09-18SPY $761.69Held: LongNext target: Long
SPY close with modeled long and short holding periods$200$300$400$500$600$700$8002020202220242026
Long held Short heldShading shows the position held during each session, after any next-open fill.

Drawdown from the prior peak

2026-09-18Model -2.44%SPY -2.08%
Drawdown from the prior peak: Risk On / Risk Off compared with holding SPY-40%-30%-20%-10%0%2020202220242026
Drawdown from the account’s previous highest daily close. The high-water mark uses the full account history, even when zoomed.

A $10,000 account following every signal from January 2, 2019 would show $87,095 at the September 18, 2026 close, against $30,954 for holding SPY over the same sessions. Both figures are price-only and exclude dividends, which understates buy-and-hold by roughly the dividend yield each year. Costs of 2 basis points per fill and 0.5% annual short borrow are already deducted.

03

The record, and why it is in-sample

The headline describes what won a search. Treat it as a description of past data rather than a forecast. Version 1 was picked out of 101,688 candidate rules by full-period annualized return over 2019-01-02 to 2026-09-17. Because the selection window runs to the freeze, the 32.52% figure below is the number that won the search. Every subperiod on this page was visible while it was being chosen.

The one exception is the forward row, which counts only sessions after the rule was frozen on 2026-09-18. It currently holds 0 sessions, so it cannot yet support any conclusion.

Selection period

2019-01-02 → 2026-09-17 · in-sample, the window the rule was chosen on

CAGR
32.52%
SPY 15.84%
Max drawdown
-23.21%
SPY -34.10%
Sharpe
1.592
SPY 0.858

Since 2023

A subperiod of the selection window

CAGR
17.12%
SPY 20.33%
Max drawdown
-23.21%
SPY -19.00%
Sharpe
1.063
SPY 1.310

Since 2025

A subperiod of the selection window

CAGR
26.25%
SPY 16.24%
Max drawdown
-13.72%
SPY -19.00%
Sharpe
1.335
SPY 0.963

Forward

Sessions after the 2026-09-18 freeze · the only out-of-sample record

No sessions yet.

CAGR and Sharpe are annualized from daily closes with a zero risk-free rate; drawdown uses daily close equity, and intraday lows are not counted. Over the selection window the model was long 75% of sessions, completed 275 trades, won 58% of them and held an average of 7.0 sessions. Since 2023 it returned 17.12% against SPY's 20.33%.

04

Year by year

Model and SPY calendar-year price returns
YearModelSPYDifferenceSessions
2019+22.60%+30.80%-8.20%252
2020+130.71%+16.16%+114.54%253
2021+24.73%+27.04%-2.30%252
2022+37.02%-19.48%+56.50%251
2023+23.91%+24.29%-0.37%250
2024-1.67%+23.30%-24.97%252
2025+44.93%+16.35%+28.58%250
2026+2.02%+11.70%-9.68%179

Calendar-year price returns, dividends excluded. The first and last rows are partial years. Years inside the selection window contributed to the rule being chosen, so a strong year there is part of the selection, not evidence against it.

05

Every trade

276 trades · newest first · open trade marked to latest close

Long SPY · Open+0.27% net
Entry open · $759.50
2026-09-16
Latest close · $761.69
2026-09-18

Signal 2026-09-15 · 3 sessions held

Short SPY+0.04% net
Entry open · $760.12
2026-09-15
Exit open · $759.50
2026-09-16

Signal 2026-09-14 · 1 sessions held

Long SPY+0.11% net
Entry open · $759.00
2026-09-14
Exit open · $760.12
2026-09-15

Signal 2026-09-11 · 1 sessions held

Short SPY+0.40% net
Entry open · $762.45
2026-09-02
Exit open · $759.00
2026-09-14

Signal 2026-09-01 · 7 sessions held

Long SPY-1.15% net
Entry open · $771.02
2026-08-07
Exit open · $762.45
2026-09-02

Signal 2026-08-06 · 18 sessions held

Short SPY+0.58% net
Entry open · $775.85
2026-08-05
Exit open · $771.02
2026-08-07

Signal 2026-08-04 · 2 sessions held

Long SPY+4.11% net
Entry open · $744.91
2026-07-27
Exit open · $775.85
2026-08-05

Signal 2026-07-24 · 7 sessions held

Short SPY-0.91% net
Entry open · $738.51
2026-07-24
Exit open · $744.91
2026-07-27

Signal 2026-07-23 · 1 sessions held

Long SPY-1.27% net
Entry open · $747.70
2026-06-22
Exit open · $738.51
2026-07-24

Signal 2026-06-18 · 23 sessions held

Short SPY-0.04% net
Entry open · $747.76
2026-06-18
Exit open · $747.70
2026-06-22

Signal 2026-06-17 · 1 sessions held

Long SPY+0.52% net
Entry open · $743.63
2026-06-09
Exit open · $747.76
2026-06-18

Signal 2026-06-08 · 7 sessions held

Short SPY-0.08% net
Entry open · $743.31
2026-06-08
Exit open · $743.63
2026-06-09

Signal 2026-06-05 · 1 sessions held

Page 1 of 23
06

Method, costs and limits

The rule

Each input maps to a value between −1 and +1. Their equal-weight average targets long above -0.15 and short below it; an exact tie or a missing input keeps the previous signal. The signal is read at a completed close and filled at the next SPY open, so the position shown at the latest close can differ from the new target. Shares stay fixed until the signal reverses, with no daily rebalancing.

Costs

2 basis points on each fill, charged on traded notional, with a reversal paying both legs. Short borrow costs 0.5% annualized on short market value over calendar days. Cash earns nothing. Dividends, interest, taxes and real broker margin requirements are not modeled, and the open trade is marked to the last close with no liquidation assumed.

What limits it

The rule was chosen by searching 101,688 combinations on this same history, which is the single largest caveat on the record above. The breadth history is a reconstruction, so point-in-time availability cannot be certified. 136 sessions since 2019-01-02 carried an incomplete input and held their previous signal rather than producing a fresh one. Latest VIX close is provisional until checked against the official settlement. Returns are price-only throughout.

How Risk On / Risk Off Works

  1. 1
    Three inputs, one fixed rule
    Each input maps to a value between −1 and +1. Their equal-weight average switches long above −0.15 and short below −0.15. An exact tie keeps the previous signal. The displayed 0–100 score puts that boundary at 42.5; it is not a probability.
  2. 2
    Act at the next session’s open
    A completed-close signal sets the next-session target. The position shown at the latest close can therefore differ from the new target. The model holds the same shares until a reversal, with no daily rebalancing.
  3. 3
    Compare with holding SPY
    The equity curves use the same SPY dates and prices. Both pay an entry cost; the long/short model also pays reversal costs and short borrow. Performance is price-only. The latest open trade is marked at the last close and excluded from completed-trade win rate.

Who Uses Risk On / Risk Off

Swing traders
Follow one daily target and inspect the signal, entry date and position history before deciding whether the rule fits your process.
Model researchers
Compare the optimized record with recent subperiods, annual returns and the separate forward record.

Pro Tips

01
Read the date before the signal
The model updates after the close. An old or incomplete reading carries its original date and should not be mistaken for a fresh market update.
02
Distinguish a target from a fill
A reversal at the latest close has not yet been filled in the historical account. The current-position panel and trade ledger show only observed fills.

Common Issues & Solutions

Risk Off while SPY is rising▾
The rule uses breadth and tape velocity contrarian. Its signal can oppose price momentum and can lose money while held.
A strong backtest but weak recent results▾
Selecting the best historical return creates selection bias. Read the dated subperiods and forward record alongside the full-period result.

Frequently Asked Questions

What does Risk On / Risk Off mean in this model?▾
Risk On means the model targets a long SPY position; Risk Off means it targets a short SPY position. It is a directional trading model. It does not measure whether markets are safe. The signal uses completed daily inputs, and a change takes effect at the next trading session’s open.
Which indicators drive the signal?▾
The model gives equal weight to the 63-session advance share, the five-session change in ES/NQ ten-session bull share, and the ten-session percentage change in VIX. Breadth and bull-share velocity are used contrarian: lower readings raise the long score. Falling VIX also raises the long score.
Are the historical returns out of sample?▾
No. Version 1 was chosen for the highest full-period annualized return in a search of one-, two- and three-indicator rules. The selection period ends September 17, 2026. The headline record and historical subperiods were visible during selection. Forward tracking is shown separately for sessions after the model was frozen on September 18, 2026.
How are entries, exits and costs modeled?▾
Signals execute at the next SPY open. Each entry uses approximately one account-equity unit of exposure, and shares remain fixed until the signal reverses. The model charges two basis points on each fill and 0.5% annual short borrow over calendar days. Dividends, interest, taxes and actual broker margin requirements are not included.
What happens if an input is missing?▾
The model retains its previous signal until all three inputs are available. The page identifies an incomplete reading and shows the last complete signal date. A signal carried through missing data is not treated as a fresh confirmation.
Is Risk On / Risk Off free?▾
Yes, during the introductory period. No account or payment is required. Paid access is planned for later; a price and transition date have not been announced.

Explore Other Tools

Advance–Decline Line
The daily breadth history behind the advance-share input.
Session Momentum
ES and NQ session control and its rate of change.
VIX Fear Index
The volatility input and its historical context.
Tape Tracker
A broader view of who controls the tape.
Last updated: 2026-09-18