giorgiofavrin.net

The Weekly Winners: An Update

Weekly Winners strategy: ten retained holdings within a twenty-stock momentum eligibility ring

The updated Weekly Winners rule keeps ten holdings while they remain within the top 20 eligible momentum ranks.

Two months into the Weekly Winners experiment, the portfolio is still ahead—but the weekly record tells a more complicated story. The gains are real, the automation is working, and the holdings are changing. The cost is very high turnover.

Weekly performance

These are actual production portfolio snapshots after each completed rebalance. Returns include the strategy’s modeled trading costs.

  • 22 July — $99.90m: −0.10% initial execution cost; −0.10% overall.
  • 20 August — $111.80m: +11.91% over four weeks; +11.80% overall.
  • 24 August — $112.61m: +0.73%; +12.61% overall.
  • 1 September — $111.63m: −0.87%; +11.63% overall.
  • 8 September — $113.12m: +1.34%; +13.12% overall, the peak so far.
  • 14 September — $108.45m: −4.13%; +8.45% overall.
  • 21 September — $109.14m: +0.64%; +9.14% overall.
  • 22 September — $108.47m: −0.61% since the latest rebalance; +8.47% overall.

The latest figure uses stored game prices updated before the regular US session, so it is the game’s marked value rather than a guaranteed live-session valuation.

What that tells us

The portfolio is up approximately $8.47 million, but most of that gain came during the first long holding period:

  • 22 July to 20 August: +11.91%
  • 20 August to 22 September, during the more regular weekly rotation: approximately −2.98%
  • From the 8 September peak: −4.11%

Operationally successful and financially positive overall—but the regular weekly-rebalancing phase has not yet demonstrated a positive edge.

Are the investments changing?

Very much so. They have not stayed in the same ten stocks:

  • 20 August: retained 3, replaced 7; added CTSH, CPRT, MSFT, SPCX, BSX, TSCO and SMCI.
  • 24 August: retained 5, replaced 5; added PLTR, ABNB, AXON, SNOW and INTU.
  • 1 September: retained 5, replaced 5; added SPCX, SLB, SNDK, VLO and HOOD.
  • 8 September: retained 4, replaced 6; added MRNA, MPC, HAL, MU, CTVA and COP.
  • 14 September: retained 5, replaced 5; added PBR, HPQ, DVN, EWZ and CMG.
  • 21 September: retained 5, replaced 5; added META, AMD, INTC, HPE and SPCX.

Across seven completed selections, 39 different stocks appeared. Normally five or six of the ten holdings changed each week. No stock appeared in every portfolio.

Current holdings: HOOD, META, VLO, AMD, MPC, INTC, HPQ, MRNA, HPE and SPCX.

Turnover forced a decision

The strategy was trading approximately half of the portfolio every week:

  • 24 August: 49.8% one-way turnover
  • 1 September: 49.4%
  • 8 September: 58.8%
  • 14 September: 49.2%
  • 21 September: 49.3%

That is roughly 51% average weekly turnover. Since inception, the experiment has generated approximately $821.5 million of gross simulated trading. At the 0.10% modeled execution penalty, that represents roughly $0.82 million of accumulated trading drag.

A stock could fall from tenth to eleventh, be sold, and then be bought back later. The momentum idea had become entangled with unnecessary boundary trading. So I changed one rule.

How the strategy works now

  • Rank eligible Wall Street Game stocks by split-adjusted return over the latest 20 completed US trading sessions.
  • Require a closing price of at least $5 and average 20-session dollar volume of at least $20 million.
  • Hold ten unique companies, approximately equally weighted, with a 2% cash reserve.
  • Retain an existing holding while it remains within the top 20 eligible momentum ranks.
  • When a holding falls below rank 20 or fails an eligibility or quote-safety check, replace it with the highest-ranked eligible candidate.
  • Keep the 0.10% modeled execution cost, live-quote safeguards and one completed rebalance per Zurich week.

What the retention buffer changes

I tested the new rule against the production portfolio and the 21 September signal data before activating it.

Under the old rule, the next selection would have sold six current holdings and bought six replacements. Under the new top-20 buffer, all ten current holdings remain eligible. Their ranks are 4, 5, 8, 10, 11, 12, 16, 18, 19 and 20.

That does not mean the portfolio will never change. It means a company must leave a wider momentum band before being sold. Strong new candidates still enter when a position drops below rank 20. The intention is to preserve momentum exposure while reducing avoidable turnover and trading drag.

The production runner has now been updated. The change itself triggered no trade and altered no historical record; it will govern the next scheduled weekly selection.

Is the automation working?

Yes. Every ISO week since the 20 August hardening has eventually completed:

  • 24 August: completed.
  • 31 August: an ABNB quote failed safely; completed on 1 September.
  • 7 September: the US market was closed for Labor Day; completed on 8 September.
  • 14 September: completed.
  • 21 September: completed.

The retry behaviour is doing what it should: a failed Monday attempt makes no trades, and the next weekday tries again.

My assessment

  • Overall result: +8.47%.
  • Regular weekly phase: approximately −2.98% so far.
  • Peak drawdown: −4.11%.
  • Historical churn: very high, usually replacing 50–60% of holdings.
  • Modeled trading drag: approximately $0.82 million.
  • Decision: keep the transparent momentum rule, but add a top-20 retention buffer.

This is a better experiment now. It still has a clear, inspectable rule, but it no longer treats tiny ranking changes as a reason to rebuild half the portfolio.

This is a simulated Wall Street Game experiment, not financial advice.

Originally published on Substack on 22 September 2026.