Part XI · Swing trading · Chapter 11G

The composite scorecard

Eight readings, one bias

1 · One number from many noisy ones

Because any single regime indicator is noisy, the composite signal aggregates eight orthogonal dimensions of market state into a single bias number, on a scale from −100 (strong bear) to +100 (strong bull). Each dimension averages several signals within it, and the composite is a weighted average across the dimensions.

It is a tilt indicator, not an entry trigger: it sets layer 1 of Chapter 11B's hierarchy, the macro regime, without replacing the invalidation rules of any trade. Its purpose is to suppress trade attempts in adverse regimes and permit them in supportive ones. In the language of Chapter 10H, it is a regime filter: it sizes trades and never makes one.

2 · The arithmetic of the score

Score each dimension $k$ from −1 (most bearish) to +1 (most bullish), and write $s_k$ for its score. If dimension $k$ has $m_k$ input signals, each already scaled to the range from −1 to +1 and written $x_{kj}$, its score is their average:

With weights $w_k$ that add up to 1, the composite signal is

Every $s_k$ lies between −1 and +1 and the weights sum to 1, so CS lies between −100 and +100. Dimension $k$ contributes $100\,w_k s_k$ points, at most $100\,w_k$ either way. That cap is also the dimension's pull on the score: raising $s_k$ by 0.1 moves CS by $10\,w_k$ points.

The framework does not say how a raw input becomes an $x_{kj}$. For illustration only, a percentile $p$ of an input that is bearish when high, such as the VIX, could be scaled as $x = 1 - 2p$. A VIX at its 80th percentile would then score $1 - 1.6 = -0.6$, and one at its 20th percentile $+0.6$.

3 · The eight dimensions

Each dimension is an independent statement about market state, from whose weighted average the composite recomputes whenever one of them changes.

Volatility and fear (18%). The VIX's IV percentile says where today's VIX sits in its own past year. The MOVE index is the bond market's counterpart of the VIX, and the SKEW index prices the third-moment risk of a crash (Chapter 5E). A high VIX percentile with a high MOVE means stress, and both are bearish components. A low VIX percentile with a low MOVE is complacency, which reads mildly bullish for mean reversion.

Credit (15%). The option-adjusted spread (OAS) is the extra yield a bond pays over Treasuries. The score reads it for high-yield bonds and for investment-grade bonds, along with its change over one month. Widening spreads mean funding stress and tightening spreads mean appetite for risk, with credit leading equity by days to weeks at major turns.

Rates (12%). The dominant input is the shape of the yield curve, the 10-year Treasury yield minus the 2-year, which is bearish over a horizon of several quarters when inverted. The 5-year breakeven inflation rate and the absolute level of the 10-year yield complete the dimension.

Dollar and FX (10%). The five-day trend of the US Dollar Index (DXY) and USD/JPY, the dollar's price in yen, carry this dimension. A rising dollar reads as risk-off, and a falling dollar as risk-on.

Commodities (8%). The inputs are the level and trend of WTI, West Texas Intermediate, the US crude oil benchmark. Gold and the ratio of the gold price to the oil price complete the dimension. A rising gold/oil ratio signals stress of a deflationary kind, while rising WTI signals inflation.

Global and overnight (10%). EEM and EFA are exchange-traded funds of emerging-market stocks and of developed markets outside the United States. With a 5-day basket of returns, they say whether the world is leading or lagging SPX. Persistent global weakness that leads SPX is bearish, while global strength with SPX flat is a bullish setup.

Trend (17%). The slope of the SPX 200-day SMA and the average directional index (ADX) of Chapter 11C carry the dimension that most retail strategies under-use. Trend up with ADX above 20 is the regime in which mean reversion is most exploitable.

Sentiment (10%). The advance-decline ratio divides the number of rising stocks by the number of falling ones. Together with the put-call ratio and the 14-day relative strength index (RSI) of SPX, it reads the crowd's mood. Sentiment is contrarian on short horizons, where extreme bullishness means pullback risk, and aligned on longer horizons, where sustained breadth means a sustained move.

4 · Treat the weights as a starting point

The weights are heuristic, so the Dealer Mechanics Lab tells its users to treat the composite as a starting framework. They should recompute it with their own weights, chosen by walk-forward backtesting, and recalibrate it periodically as regimes shift.

The score is most informative as a directional indicator: regime strongly supportive, weakly supportive or hostile. It is less informative as a precise value, so anchor decisions on the regime bucket and treat the numerical score as a confidence indicator.

Fitting eight weights is harder than it looks, because the data contain only a handful of true regime changes. Eight free weights can fit those few episodes almost perfectly and learn nothing from them (overfitting, Chapter 3A). Chapter 1D's multiple-comparisons problem applies to every set of weights you try. The honest test is out of sample: choose the weights on one period and judge them on a later one.

A cheaper check than a full backtest is a sensitivity test on the weights you already have. Move five points of weight from trend to sentiment in the worked example, and the composite falls by $5 \times (0.6 - (-0.4)) = 5.0$ points to +12.4, which is still a bull tilt. A conclusion that survives such shifts is a property of the market rather than of your weights.

5 · From score to action

As the macro-layer input to the hierarchy, the composite maps to action through five buckets:

Each edge belongs to the more extreme bucket: +10 and +40 are on the bullish side of their edges, and −10 and −40 on the bearish side. The foundation sleeves of the strong-bear row are mechanical sleeves holding the risk premia of Chapter 10A: they follow their own rules and are not stopped.

6 · The scorecard

The figure gives each dimension a slider from −1 to +1, in steps of 0.05, all starting at +0.20. The top panel draws each score times its weight, in points of the composite, over a pale track as wide as that weight allows. The lower panel places the composite on the scale from −100 to +100, whose shaded bands are the five buckets, and a readout names the bucket and its action.

The composite is shown to one decimal, cut toward zero rather than rounded. The bucket is chosen on the exact value, so the number never reads as past an edge its bucket has not crossed. Start at the defaults, then set every slider to zero except trend at +0.35 and commodities at +0.50.

volatility / fear

+0.20

credit

+0.20

rates

+0.20

dollar / FX

+0.20

commodities

+0.20

global / overnight

+0.20

trend

+0.20

sentiment

+0.20

Bull tilt, composite +20.0. Long trades permitted at 70-80% of intended size.

At the defaults every dimension contributes a fifth of its maximum. The readout is "Bull tilt, composite +20.0", with long trades at 70% to 80% of intended size. Drag volatility alone to −1 and the composite falls by $1.2 \times 18 = 21.6$ points to −1.6, a neutral regime.

With only trend and commodities set, the composite is $17 \times 0.35 + 8 \times 0.50 = 9.95$, and it reads "Neutral, composite +9.9". Rounded, it would read +10.0 beside the word neutral, which is the confusion the cut toward zero avoids. One more step of commodities, to +0.55, gives 10.35 and a bull tilt. With every other slider at zero, setting dollar and FX to −1 gives exactly −10.0, which reads as a bear tilt, since the edge belongs to the bearish side.

7 · The composite and the conviction score

The composite answers the question of what the regime is. The conviction score of Chapter 13B answers a different one: given the regime, should you take this specific trade? The two are complementary. A composite of +50 does not earn you the trade, because you still need to score the product $E \times M \times R \times X \times S$ of evidence, mechanism, regime, execution and sizing. But a composite of −60 zeroes the regime factor $R$ for most long strategies, however high the other factors are.

Since the conviction score is a product, a zero anywhere makes it zero. With Chapter 13B's calm-regime scores of $E = 0.85$, $M = 0.90$, $X = 0.90$ and $S = 0.90$, a regime factor of 0.85 gives a conviction of 0.527. With $R = 0$ the same trade scores 0, however strong its other factors.

Use the composite to filter the universe of trades you consider. Use the conviction score to evaluate the trades that survived the filter.

8 · The Edge Book builds the same layer differently

The Edge Book sizes its short-volatility sleeves with a stack of a few filters rather than a weighted average (Chapter 10H). Full exposure needs every condition green: the VIX/VIX3M ratio below 1.00, 21-day realized volatility below implied, and the TED spread, a gauge of banks' funding stress, below 100 bps. The MOVE index must be below 100 and high-yield OAS below 500 bps, with positive gamma exposure as a low-confidence supplement. With all green it runs full size, while one yellow reduces size by 25% and two or more reduce it by 50% or more.

The Edge Book also names what is not a filter, starting with single oscillator readings such as an RSI at 30, which are too noisy for sizing decisions. Sentiment indicators, it says, are noisy and fit for contrarian use only. A filter, it argues, needs continuous measurement, demonstrated predictive value for the regime and a basis that does not rest on a story. The composite's sentiment dimension leans on inputs of exactly those kinds: an oscillator, the 14-day RSI, and a reading of the crowd, the put/call ratio. The two sources agree that the regime should size trades. They differ on how to combine the inputs, and on which inputs deserve a say.

9 · Check yourself