Skip to content

14 Comments

  1. Lenn
    November 7, 2021 @ 10:06 am

    Hello,

    You say that
    “Performance of the s1 signal is explored in different contexts in other, non-paywalled s1 signal studies. In particular, different durations (0-3 DTE, 7 DTE, 45 DTE and LEAPS), different strategies (uncovered and verticals)…”

    However I could only find the two posts comparing put spreads to short puts, not any Posts comparing different DTEs.
    Is there any way to see a comparison between the different DTEs? I’d be really interested.

    Best

    Reply

    • spintwig.com
      November 7, 2021 @ 12:08 pm

      Yes! The 45 DTE version the short put study is being released this Friday. I’m also pulling together the data for the other DTEs and approaches as well. Stay tuned!

      Reply

  2. Steve Highland
    November 12, 2021 @ 11:23 am

    11-12-2021
    Hi, This may be obvious but I’m new here. I’m trying to understand what the ‘s1 signal’ is all about. I have read here: “The s1 signal is a boolean (TRUE / FALSE) daily indicator that attempts to identify the days in which short put and short vertical put positions on the S&P 500 are most likely to be profitable at expiration. s1 is based on data from CBOE and S&P Global.”
    I take it from this that it is a quant strategy based on TECHNICAL indicators, right? I realize it must be proprietary but can you please give more information on what it is and why it is likely to be reliable in order to help me understand what it is and why I should explore it further?
    Thanks!
    Steve

    Reply

    • spintwig.com
      November 20, 2021 @ 5:44 pm

      Welcome Steve — correct, the s1 signal is quantitive in nature and uses historic values to suggest when to open a position and when to sit out.

      It is reliable for several reasons:

      1) s1 uses market data as inputs – simply stated: relevant, empirical data is used. There are no “moon phases” or other nonsensical inputs.

      2) s1 data inputs are from official sources – using CBOE and S&P Global as the data providers ensures input data is as clean as possible and free from influence or error of 3rd parties

      3) s1 avoids overfitting – it’s easy to look back in time and identify an “optimal” methodology that avoids all the drawdowns. The problem with this approach is that it tunes the algo to both the signal as well as the noise in the market. The result is typically a “perfect” backtest and a poor forward-looking result. The s1 signal paints with “broad strokes” and doesn’t look for a “perfect” backtest. The result is typically a “great” backtest and a great forward-looking result.

      Hope this helps!

      Reply

  3. Dom
    November 12, 2021 @ 12:27 pm

    I’m not too sure I see the point in this, for the effort involved and time spent! I could have a 98% cash and 2% BTC portfolio that would outperform the standard buy and hold and any of these active strategies.

    You’re doing all this and not even really generating any Alpha and with a 6%+ drag from inflation right now you’re doing all this for a poultry <4% a year… kinda pathetic really.

    I think all you've proved here is that taking on the extra risk just isn't worth the returns involved.

    Reply

    • spintwig.com
      November 20, 2021 @ 5:50 pm

      I’m not too sure I see the point in this, for the effort involved and time spent! I could have a 100% BTC and 0% cash portfolio that would outperform your proposed strategy as well as buy and hold and any of these active strategies.

      Anyone can look back and say what would have made more/less money. I’m not sure I see your point.

      Reply

  4. M
    December 21, 2021 @ 11:16 am

    I appreciate the need to be vague on what kind of data the S1 signal uses but is there any details you can give? Is it some sort of price mean reversion thing?

    Reply

    • spintwig.com
      December 23, 2021 @ 12:48 pm

      I can share that it uses official data sources from CBOE and S&P Global as inputs. However, I suspect that’s less than helpful for what you’re looking for.

      As for the mechanics, there’s not much I can say without risking someone reverse engineering the model. There are a lot of smart folks out there! I can say that it’s a systematic and repeatable calculation; there are no chances that a historical signal value would ever change from true to false or vice versa.

      Reply

  5. Adam
    January 23, 2022 @ 2:36 am

    Hi

    Nice work! I’m just trying to understanding this … The tail end of your research discussed 10D-2.5D put credit spreads, even though the 50D-5D produced the highest profit. Does this mean the 10D-2.5D put credit spread (S1 signal) is the wisest choice to trade, with all other factors considered?

    Reply

    • spintwig.com
      January 26, 2022 @ 1:39 am

      Thanks Adam!

      My selection of the 10D-2.5D strategy in the discussion section was a random choice among the available strats in the study. There is nothing inherently special about it over another configuration.

      As for which is the wisest strategy to trade, that would be up to you. I’d suggest identifying one or more specific performance goals (eg: “must have a CAGR of no less than 6%”, or “must have a Sharpe ratio of no less than 1.00” or “must have a max drawdown no garter than ‘x’ percent”) that you’d like to achieve, then implement the strat that performs closest to those goals. Your goals may be similar or different vs others’ goals.

      Reply

  6. Mario
    March 17, 2022 @ 6:35 am

    Hi

    I hope this is not a stupid question, how was starting capital calculated?

    Looking at all the info in the backtest everything else looks pretty similar (nr of concurrent positions, notional exposure, margin utilization, etc). Why is the starting capital different for Daily entry and signal entry?

    Reply

    • spintwig.com
      March 17, 2022 @ 10:21 am

      Hey Mario – not a stupid question at all!

      The starting capital is different because of some of the backtests perform better than others.

      Suppose an arbitrary, fixed amount of capital is assigned at the beginning of each backtest. The better-performing strategies may not need all the capital. They may be able to execute the strategy with, say, 20% less money. Similarly, underperforming strategies may get blown out with margin calls and the strategy can’t run for the entire duration of the backtest. More capital may be needed.

      In order to solve for this and accurately represent the performance across strategies of differing performance, I normalize everything to a maximum margin utilization of 100%. In other words, I find the minimum amount of capital needed to successfully execute the strategies over the entire duration of the backtest and set that as the starting capital. It allows the purest representation of strategy performance – no cash drag for the outperforming strategies and no margin calls for the underperforming strategies.

      Despite the differing stating amounts, CAGR (geometric return) is calculated based on starting and ending portfolio values and dates – the standard way to calculate CAGR. This ensures that despite the different amounts of starting capital, strategy performance is accurately calculated across strategies.

      Reply

  7. Mario
    March 17, 2022 @ 10:39 am

    Hi, thanks for the fast response.

    I guess that makes sense but at the same time, I don’t understand the difference. With a similar number of winning trades, similar concurrent positions, etc, why is the strategy without S1 needing double the capital?

    Reply

    • spintwig.com
      March 18, 2022 @ 1:25 am

      It mostly has to do with the loss frequency and severity, and a bit of sequence of returns risk.

      Consider the “win rate statistics” table in the discussion section. The loss rate for the “all” strategy is 4.80% while the loss rate for the s1 = true strat is 4.19%. The s1 = true strat experienced 12.7% fewer losing trades.

      It’s not uncommon for a trader to experience hundreds of profitable trades only to experience a single losing trade that wipes out months (or years) of profit. By reducing the occurrences of losing trades by just 12.7%, the strategies experience a significant outperformance.

      As for sequence risk, if a strategy experiences a material loss right out of the gate, this can disproportionally hurt the strategy vs experiencing a material loss later in the backtest. This wasn’t a major factor in this scenario though, as demonstrated by the max margin utilization dates.

      Reply

Leave a Reply

Your email address will not be published. Required fields are marked *