Nils W

BOTW 25: Robots to the Rescue Exploiting potential rebalancing candidates among the S&P sector ETFs produces nice risk-adjusted outperformance

𝗕𝗢𝗧𝗪 𝟮𝟱: 𝗥𝗼𝗯𝗼𝘁𝘀 𝘁𝗼 𝘁𝗵𝗲 𝗥𝗲𝘀𝗰𝘂𝗲: New backtest inspired by Robot Wealth. Go long the bottom half performing S&P sector ETFs from the first half of the month and hold until the end of the month. Risk-adjusted returns exceed SPY for time in the market.

optionstocksmachines.substack.com/p/botw-25-ro...

BOTW 16: Oil vs. Oil Pairs trading big oil companies Exxon or Chevron vs. oil significantly outperforms XLE and the S&P.

𝗕𝗢𝗧𝗪 𝟭𝟲: 𝗢𝗶𝗹 𝘃𝘀. 𝗢𝗶𝗹 Pairs trading big oil companies vs. oil the commodity significantly outperforms XLE and the S&P. But a missing component could alter results meaningfully. Stay tuned for the reveal in our next post!

optionstocksmachines.substack.com/p/botw-16-oi...

HFW 19: Automate ETS AutoETS improves forecasts meaningfully vs. ETS on Apple

𝗛𝗙𝗪 𝟭𝟵: 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲 𝗘𝗧𝗦 We compare an additive ETS model against an AutoETS one on Apple's revenues. Forecast errors meaningfully lower for AutoETS. Warrants optimism around AutoETS performance when we test it on the full dataset. Stay tuned!
optionstocksmachines.substack.com/p/hfw-19-aut...

optionstocksmachines.substack.com/p/botw-15-ey... Recently, while private credit has become investable for individual investors, it hasn't always been. We take a crack at isolating returns to the asset over a 10-year period in recent backtest.

optionstocksmachines.substack.com/p/botw-15-ey...

HFW 18: Adding the E to ETS Adding an error term to our exponential smoothing trend and seasonality models produces modestly better performance than without the term

𝗛𝗙𝗪 𝟭𝟴: 𝗔𝗱𝗱𝗶𝗻𝗴 𝘁𝗵𝗲 𝗘 𝘁𝗼 𝗘𝗧𝗦 Adding an error term to the exponential smoothing models we use to forecast revenues on our S&P 500 sector dataset improves performance modestly. This sets up nicely to introduce AutoETS in future posts. Stay tuned!
optionstocksmachines.substack.com/p/hfw-18-add...

BOTW 14: Systematic Factors A long/short portfolio created by ranking factor ETFs using rolling Sharpe Ratios beats the market.

𝗕𝗢𝗧𝗪 𝟭𝟰: 𝗦𝘆𝘀𝘁𝗲𝗺𝗮𝘁𝗶𝗰 𝗙𝗮𝗰𝘁𝗼𝗿𝘀 We take a breather from the rarefied air of dividend aristocrats to wade in on factor investing for our weekly backtest. A long/short portfolio created by ranking factor ETFs using rolling Sharpe Ratios beats the market.

optionstocksmachines.substack.com/p/botw-14-sy...

HFW 17: Smooth Trends Exponential smoothing with trend and seasonality performs as well as simple smoothing but with more realistic forecast graphs

𝗛𝗙𝗪 𝟭𝟳: 𝗦𝗺𝗼𝗼𝘁𝗵 𝗧𝗿𝗲𝗻𝗱𝘀 Exponential smoothing with trend and seasonality performs as well as simple smoothing but with more realistic forecast graphs, a key enhancement for stakeholders that need such details. More enhancements to follow. Stay tuned!

optionstocksmachines.substack.com/p/hfw-17-smo...

Agreed! Thanks for sharing.

BOTW 13: Doggie Deux Using deciles instead of a top 10 ranking improves our Aristocratic Dog strategy by 10 percentage points.

𝗕𝗢𝗧𝗪 𝟭𝟯: 𝗗𝗼𝗴𝗴𝗶𝗲 𝗗𝗲𝘂𝘅 More Dogs of the Dow with a Twist! Long-short strategy on the top and bottom deciles by dividend yield outperforms by over 50% pts vs benchmark. Second from the top and bottom underperforms by about 10% pts. Real doggie deux!

optionstocksmachines.substack.com/p/botw-13-do...

HFW 16: Smooth Exponentialer Simple exponential smoothing yields the best performance thus far

𝗛𝗙𝗪 𝟭𝟲: 𝗦𝗺𝗼𝗼𝘁𝗵 𝗘𝘅𝗽𝗼𝗻𝗲𝗻𝘁𝗶𝗮𝗹𝗲𝗿 Simple exponential smoothing outperforms the benchmarks on our S&P 500 revenue dataset. But the model produces flat forecasts just like the naïve one. We'll resolve that issue in our next post. Stay tuned!

optionstocksmachines.substack.com/p/hfw-16-smo...

BOTW 12: Aristocratic Dogs Applying the Dogs of the Dow methodology to Dividend Aristocrats yields significant outperformance relative to the benchmark

𝗕𝗢𝗧𝗪 𝟭𝟮: 𝗔𝗿𝗶𝘀𝘁𝗼𝗰𝗿𝗮𝘁𝗶𝗰 𝗗𝗼𝗴𝘀 Not King Charles Spaniels or the new Tarantino movie. Applying the Dogs of the Dow strategy to the Dividend Aristocrats. Results are impressive. Beats the benchmark by almost 40% pts. Quality investing for quants!
optionstocksmachines.substack.com/p/botw-12-ar...

BOTW 11: Credit Spread Arbitrage A pairs trading strategy using the ETFs LQD and HYG generates a Sharpe Ratio 20% better than an equivalent 50-50 benchmark on the same ETFs

𝗕𝗢𝗧𝗪 𝟭𝟭: 𝗖𝗿𝗲𝗱𝗶𝘁 𝗦𝗽𝗿𝗲𝗮𝗱 𝗔𝗿𝗯𝗶𝘁𝗿𝗮𝗴𝗲 Continuing our fixed income theme, a pairs trading strategy using the ETFs LQD and HYG generates a Sharpe Ratio 20% better than an equivalent 50-50 benchmark. More fixed income backtests to follow. Stay tuned!
optionstocksmachines.substack.com/p/botw-11-cr...

I feel there's a wonderful analogy waiting to emerge...

HFW 14: GARCH Curious G/ARCH models perform better than most models examined to date. But our time series are probably not sufficiently long to allow these models to shine.

𝗛𝗙𝗪 𝟭𝟰: 𝗚𝗔𝗥𝗖𝗛 𝗖𝘂𝗿𝗶𝗼𝘂𝘀 Key takeaway: G/ARCH models beat most models examined to date in our Hello Forecasting World series. But the naïve and autoregressive growth rate models remain at the top. Simplicity wins the day, again!

optionstocksmachines.substack.com/p/hfw-14-gar...

BOTW 10: MBS for Fun and Profit Pairs trading MTBA vs. MBB, the MBS ETFs, generates tidy risk-adjusted outperformance with or without leverage

𝗕𝗢𝗧𝗪 𝟭𝟬: 𝗠𝗕𝗦 𝗳𝗼𝗿 𝗙𝘂𝗻 𝗮𝗻𝗱 𝗣𝗿𝗼𝗳𝗶𝘁 We pair two MBS ETFs, MBB and MTBA. The strategy outperforms nicely, and, generates a Sharpe Ratio double the benchmark. Leverage warranted and while remaining market neutral.
optionstocksmachines.substack.com/p/botw-10-mb...

HFW 13: ARCH Complex A basic ARCH model performs better than most of the previous models, but is modestly worse than the naïve and autoregressive growth rate models

𝗛𝗙𝗪 𝟭𝟯: 𝗔𝗥𝗖𝗛 𝗖𝗼𝗺𝗽𝗹𝗲𝘅 Addressing volatility in S&P 500 revenue dataset with a basic ARCH model. Bottom line: ARCH beats most others to date apart from the benchmarks. Performs the best on the tech ETF, XLK, More knobs to tune performance!
optionstocksmachines.substack.com/p/hfw-13-arc...

Recession indicators: Sahm Rule I Not a forecasting tool, but no false positives like yield spreads

𝗥𝗲𝗰𝗲𝘀𝘀𝗶𝗼𝗻 𝗜𝗻𝗱𝗶𝗰𝗮𝘁𝗼𝗿𝘀: 𝗦𝗮𝗵𝗺 𝗥𝘂𝗹𝗲 𝗜 Key finding: though not a forecasting tool, the rule does not suffer from the same type of false positives one sees in yield spreads. Our next post will compare the rule to stock market action. Stay tuned!

optionstocksmachines.substack.com/p/recession-...

HFW 12: ARMA adds little again An ARMA(1,1) model using growth rates performs modestly worse than the AR(1) model, but still better than most of the prior models we've tested thus far

𝗛𝗙𝗪 𝟭𝟮: 𝗔𝗥𝗠𝗔 𝗮𝗱𝗱𝘀 𝗹𝗶𝘁𝘁𝗹𝗲 𝗮𝗴𝗮𝗶𝗻 Adding a moving average component does little for performance. But the model does perform better than all the previous ones apart the benchmarks. Time to move on to incorporating volatility into our models. Stay tuned!
optionstocksmachines.substack.com/p/hfw-12-arm...

BOTW 8: Rolling with AAII Rolling regression on AAII signals improves strategy results meaningfully

𝗕𝗢𝗧𝗪 𝟴: 𝗥𝗼𝗹𝗹𝗶𝗻𝗴 𝘄𝗶𝘁𝗵 𝗔𝗔𝗜𝗜: Rolling regression for the win. We build several rolling regression models, rank them, and then satisfice, rather than optimize. The resulting outperformance on the validation set exceeds 30% points!
optionstocksmachines.substack.com/p/botw-8-rol...

HFW 11: Autoregressive growth rates An AR(1) model using sequential growth rates takes the title from the naïve model for best performance on our S&P dataset

𝗛𝗙𝗪 𝟭𝟭: 𝗔𝘂𝘁𝗼𝗿𝗲𝗴𝗿𝗲𝘀𝘀𝗶𝘃𝗲 𝗴𝗿𝗼𝘄𝘁𝗵 𝗿𝗮𝘁𝗲𝘀 An AR(1) model using growth rates surpasses the major algorithms we've examined thus far, and is also tiny bit better than the naïve model. Next up we'll see if we can improve upon this performance. Stay tuned!
optionstocksmachines.substack.com/p/hfw-11-aut...

But there's still a 30% probability of negative returns. On a 6-month horizon, the upside exceeds the down by about 18 percentage points. Not bad!

The 200-Week Empirical probabilities suggest the best risk-reward comes in the 6-month period after the S&P 500 dips below the 200-Week moving average

𝗧𝗵𝗲 𝟮𝟬𝟬-𝗪𝗲𝗲𝗸 Wall St. strategists are saying the S&P might find support at the 200-Week Moving Average. Classic data analysis shows one-month to one-year returns after a dip below tend to be positive.
optionstocksmachines.substack.com/p/the-200-week

BOTW 7: Momentum of momentum A pairs strategy using momentum as a signal on momentum and covered call ETFs produces attractive outperformance vs. the benchmark

𝗕𝗢𝗧𝗪 𝟳: 𝗠𝗼𝗺𝗲𝗻𝘁𝘂𝗺 𝗼𝗳 𝗺𝗼𝗺𝗲𝗻𝘁𝘂𝗺 Pairing the S&P momentum ETF against the covered call variant produces over 50 percentage points of cumulative outperformance vs. the benchmark. Definitely an attractive proof-of-concept. Stay tuned!
optionstocksmachines.substack.com/p/botw-7-mom...

Intraday Momentum Limited evidence of the momentum effect on 5-minute bars in the SPY

𝗜𝗻𝘁𝗿𝗮𝗱𝗮𝘆 𝗠𝗼𝗺𝗲𝗻𝘁𝘂𝗺 Momentum may be the premier anomaly. But it doesn't seem to scale fractally. Analyzing 5-minute SPY ETF bars shows limited evidence of the momentum effect. More to do including adjusting for overnight returns. Stay tuned!
optionstocksmachines.substack.com/p/intraday-m...

HFW 10: When average wins An average growth rate forecast model beats the more sophisticated models thus far

𝗛𝗙𝗪 𝟭𝟬: 𝗪𝗵𝗲𝗻 𝗮𝘃𝗲𝗿𝗮𝗴𝗲 𝘄𝗶𝗻𝘀 Using historical average growth rate with forward recursion outperforms all previous models, except, original naïve benchmark. But try recommending a naïve model to stakeholders. Heteroskedasticity models next. Stay tuned!
optionstocksmachines.substack.com/p/hfw-10-whe...

Crypto Regimes Hierarchical clustering generates attractive trading signals in Bitcoin and Ethereum

𝗖𝗿𝘆𝗽𝘁𝗼 𝗥𝗲𝗴𝗶𝗺𝗲: Revisiting our work on regime detection, we use clustering algorithms on crypto to generate trading signals, producing powerful outperformance. Stay tuned for other regime detection algorithms in upcoming posts.
optionstocksmachines.substack.com/p/crypto-reg...
#python #datascience

BOTW 6: AAII Redux We build a relatively stable model using market implied reactions to the AAII survey

𝗕𝗢𝗧𝗪 𝟲: 𝗔𝗔𝗜𝗜 𝗥𝗲𝗱𝘂𝘅 Using prediction levels to drive trading signals we develop a strategy that outperforms buy-and-hold on cumulative and risk-adjusted returns. Not bad for a relatively straightforward model. Stay tuned for additional refinements!
optionstocksmachines.substack.com/p/botw-6-aai...

HFW 9: Growth rate forecasting Setting the stage for volatility clustering

𝗛𝗙𝗪 𝟵: 𝗚𝗿𝗼𝘄𝘁𝗵 𝗿𝗮𝘁𝗲 𝗳𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴 Adding growth rate forecasting to toolbox. Provides intuition on 𝘴𝘵𝘢𝘵𝘪𝘰𝘯𝘢𝘳𝘪𝘵𝘺, but also motivates examination of next set models that explicitly forecast volatility. Stay tuned!
optionstocksmachines.substack.com/p/hfw-9-grow...
#python #datascience #timeseries

BOTW 5: Covered call ETFs Nice idea, but they don't perform as well as expected. A long-short portfolio of actively vs. passively managed ETFs looks promising though.

𝗕𝗢𝗧𝗪 𝟱: 𝗖𝗼𝘃𝗲𝗿𝗲𝗱 𝗖𝗮𝗹𝗹 𝗘𝗧𝗙𝘀 Perform poorly on a cumulative and risk-adjusted return basis. Long-short active-passive covered call ETF portfolio enjoys substantial risk-adjusted outperformance. Stay tuned as we drill down into this in further posts! optionstocksmachines.substack.com/p/botw-5-cov...

One Naïve Portfolio RSP does not meaningfully outperform SPY

𝗢𝗻𝗲 𝗡𝗮ï𝘃𝗲 𝗣𝗼𝗿𝘁𝗳𝗼𝗹𝗶𝗼 Over the long term, equal-weighted RSP performs similar to the SPY for a higher expense ratio. A tactical allocation might be warranted. Need to confirm in a follow-on post. Stay tuned!
optionstocksmachines.substack.com/p/one-naive-...
#python #datascience #workingidea