Start of live Nasdaq-100 strategy testing: 1993-10-01
Momentum lookback: 252 trading days
Base rebalance schedule for the reposted core results: first trading day of each calendar month
Friction: 0.1% of traded notional at each rebalance
Timing filter: hold cash for the next period whenever SPY is below its 325-day moving average on the rebalance date
Beta and Treynor are measured versus SPY daily returns
Sharpe and Sortino use 0% risk-free rate

momentum
look
back Top5 CAGR	Sharpe	Max DD	5timed timed Shrp timed Max DD
3M	21.1%	0.708	-81.9%	19.1%	0.722	-53.7%
6M	23.0%	0.742	-81.8%	25.0%	0.847	-58.9%
9M	25.8%	0.789	-70.2%	27.1%	0.870	-55.6%
12M	23.8%	0.742	-69.7%	23.6%	0.790	-50.7%
18M	13.6%	0.518	-87.3%	17.5%	0.640	-63.2%

So 9 month momentum look back for highest CAGR; 12 month gave a bit better on max DD.

For holding periods 1 month was best:

Hold	5 CAGR	5 Shrp	5 MaxDD	5timed 5timedShrp 5timed Max DD
2 weeks	21.2%	0.690	-78.9%	25.8%	0.859	-54.3%
1 month	23.8%	0.742	-69.7%	23.6%	0.790	-50.7%
2 month	17.7%	0.611	-82.4%	19.3%	0.686	-50.8%
3 month	18.4%	0.621	-88.8%	19.7%	0.686	-67.3%
6 month	13.4%	0.515	-83.3%	9.5%	0.432	-77.3%

For the continuous monthly-schedule runs, profit concentration looked like this:

Scenario
Top 1% of winning positions Top 5% 10%
Mom 1 4.9% 20.6% 35.7%
Mom 1 timed 5.5% 22.4% 37.2%
Mom 5 8.0% 24.3% 36.6%
Mom 5 timed 8.1% 24.3% 37.1%
Mom 10 9.4% 27.0% 39.3%
Mom 10 timed 9.1% 26.7% 39.2%

The strategy does not look like “one freak winner carries everything.” Missing 1 wouldn't destroy your return, but missing more of the top winners would. Missing 10% of your best winners would hurt more.

A 10-stock strategy with the timing rule is the best on 5 and 10 year horizons of all the versions, beating the S&P over the worst 2,3,5, and 10 year stretches, and coming close on the worst 1 and 2 year stretches for the S&P500 timed version itself:
What if someone started at a bad (the single worst) time?:

Worst rolling CAGR by horizon:

Scenario	Worst 1Y  2Y	  3Y	  5Y	10Y
NDX Mom 1	-70.3%	-42.1%	-28.3%	-12.9% -1.0%
Mom 1timed	-48.6%	-27.3%	-18.0%	- 2.3%	4.4%
NDX Mom 5	-46.0%	-20.8%	-12.6%	- 2.4%	2.8%
Mom 5timed	-28.4%	-12.3%	- 6.3%	  0.8%	5.0%
Mom 10	        -38.7%	-17.0%	- 9.5%	- 0.8%	2.5%
Mom 10timed	-25.5%	-10.2%	- 4.8%	  1.8%	5.1%
SPY	        -43.3%	-23.3%	-14.5%	- 2.3%	1.2%
SPYtimed	-22.0%	- 7.8%	- 2.7%	  2.3%	4.9%
QQQ	        -67.6%	-38.0%	-25.0%	- 9.4%	0.7%
QQQtimed	-35.0%	-16.1%	- 9.0%	- 0.7%	3.8%

First-trading-day-of-month start sensitivity:
This is the “what if I happened to begin on the first trading day of any month?” test:

Scenario
# Starts AvgStartCAGR Worst AvgShrpe AvgSortino AvgBeta AvgTreynor Avg Start Max DD
Mom 1 378 22.9% -1.5% 0.674 0.951 1.383 0.179 -78.7%
Mom 1timed 378 28.0% 3.8% 0.761 0.938 1.096 0.284 -70.2%
Mom 5 378 18.8% 3.9% 0.612 0.848 1.170 0.165 -55.4%
Mom 5timed 378 18.2% 6.4% 0.663 0.793 0.799 0.229 -44.1%
Mom 10 378 16.8% 3.3% 0.586 0.791 1.110 0.156 -50.6%
Mom 10timed 378 16.9% 7.1% 0.653 0.759 0.775 0.223 -39.9%
SPY 378 11.3% 4.1% 0.661 0.859 0.924 0.123 -41.8%
SPYtimed 378 9.8% 4.9% 0.699 0.776 0.565 0.177 -31.3%
QQQ 312 11.3% -0.5% 0.526 0.699 0.964 0.129 -42.3%
QQQtimed 312 11.0% 3.8% 0.613 0.701 0.617 0.188 -28.4%What that says: The top-1 version is still the best return engine, but starting-month luck matters a lot.
The timing filter materially improved the worst starting experience.
Top-10 +325DMA looks like the most forgiving momentum version by average drawdown

Core continuous results from the earliest start:

Scenario	CAGR	Vol	Sharpe	Sortino	Beta	Treynor	Max DD	Avg Turnover
Mom 1	        33.7%	65.6%	0.780	1.121	1.544	0.322	-81.2%	83.3%
Mom 1timed	37.0%	56.9%	0.855	1.082	1.250	0.471	-76.0%	64.9%
Mom 5	        23.8%	40.0%	0.742	1.041	1.329	0.219	-69.7%	70.6%
Mom 5timed	23.6%	33.5%	0.790	0.965	0.885	0.306	-50.7%	58.1%
Mom 10	        21.2%	34.7%	0.735	1.002	1.267	0.198	-71.6%	62.0%
Mom 10timed	21.8%	29.4%	0.813	0.962	0.887	0.288	-48.0%	50.9%
SPY	        10.5%	18.7%	0.626	0.798	0.892	0.117	-55.2%	 0.3%
SPYtimed	 9.4%	14.0%	0.706	0.783	0.541	0.174	-33.7%	 6.4%
QQQ	        10.0%	24.6%	0.488	0.645	0.900	0.111	-83.0%	 0.3%
QQQtimed	10.9%	18.0%	0.621	0.704	0.567	0.192	-46.6%	 5.9%

My read from that set:

Best raw return: NDX Mom 1 +325DMA
Best balanced momentum version: NDX Mom 5 +325DMA or NDX Mom 10 +325DMA
Best plain ETF drawdown control: SPY +325DMA
The 325DMA filter helped drawdowns a lot, but did not always improve CAGR.