tradehorde

Building a Market Forecasting Engine that keeps score.

[tradehorde @tradehorde]

September update: 20 public signals resolved as wins or losses.⚠️Win rate: 45.0%

September’s 50+ cohort: 18 signals, ✅50.0% wins, ✅+0.12R average and ✅1.50 profit factor.

@tradehorde’s record since inception now includes 359 win/loss-resolved public signals, including 192 with conviction scores of 50+.

All-time stats for signals with score > 50:
✅55.7% win rate — 107W / 85L
✅+0.57R average outcome per signal
✅2.31 profit factor
✅Winners: +1.72R / Losers: −0.88R

A weaker month, included in the record.
Results span multiple models and system versions. These are hypothetical signal outcomes before trading costs.

[tradehorde @tradehorde]

Mid-month Update.
August-to-date production Brier is 0.2096✅
Win rate: 66.7%✅

@tradehorde forecasting record since inception.
328 resolved public signals out of which 163 with 50+ conviction.

Below are stats for the 163 with conviction score 50+
✅56.4% win rate
✅+0.63R average return
✅ 2.50 Profit Factor
✅Winners: +1.8R average / Losers: −0.9R average

[tradehorde @tradehorde]

$BTC successful retest of 84K here could mean rapid expansion.

I still think it all resolves by the 30Y auction on Thursday.

Tomorrow is 10Y auction and FOMC minutes.

So lots of whiplash incoming around yields.

[tradehorde @tradehorde]

$BTC travel time from 84K to 94K has been fast, in the past.

Median is ~ 2.6 days after a successful 84K retest.

[tradehorde @tradehorde]

Crypto taking a crap.

[tradehorde @tradehorde]

The longer term trends favor $BTC upward continuation.
Shorter-timeframe is very muddy.

Wed & Thursday have some bond auctions that might eventually produce the big moves.

[tradehorde @tradehorde]

The longer term trends favor $BTC upward continuation.
Shorter-timeframe is very muddy.

Wed & Thursday have some bond auctions that might eventually produce the big moves.

[tradehorde @tradehorde]

🤣

[Paulo Macro @PauloMacro]

Calling this plague outcome now - oil negative, stocks +50%... just imagine how many GPUs we need to fix this...

[tradehorde @tradehorde]

when more ai models than meme coins?

[tradehorde @tradehorde]

After the ISM data earlier and last weeks data - Fed hike odds in Oct are minimal.
That still leaves the long end which has several events this week Wed & Thursday with 10Y/30Y auctions and FOMC minutes on Wed.
$SPY continues to be resilient.

[tradehorde @tradehorde]

This week's US Research Desk comes down to Yen Carry Squeeze into the October BOJ and AI-Driven Memory Supercycle.
Week of Oct 5.

1. Yen Carry Squeeze into the October BOJ
With the Fed widely expected to skip October (CME-implied 17% hike odds after the soft jobs report), the differential is more likely to compress than widen into month-end, squeezing crowded short-yen carry.
$USDJPY · bearish · swing
Theme invalidation: The theme fails if USD/JPY records a weekly close above 160.08 while the U.S.-Japan two-year yield spread rises above 3.0 percentage points.
USD/JPY invalidation: A weekly close above 160 with no intervention and widening US-Japan rate spreads.

2. AI-Driven Memory Supercycle
The cleanest US expression is the only large-cap memory pure play, with monthly contract-price prints and quarterly earnings as dated monitoring milestones across the position window.
MU / SNDK · bullish · position
Theme invalidation: The theme breaks if monthly TrendForce contract price prints turn flat-to-negative, or if Samsung/SK hynix announce conventional-DRAM capacity additions that reverse the 2027-2028 shortfall guidance, or if hyperscaler AI capex guidance is cut.
MU invalidation: A sequential decline in reported DRAM/NAND contract pricing or a Micron downward revision to HBM 2027 volumes/pricing.
SNDK invalidation: NAND spot/contract prices rolling over or SNDK guidance signaling demand softness.

Research hypotheses, not active signals.

Full analysis:
https://t.co/LgM0LjCZqH

[tradehorde @tradehorde]

The public signal scorecard, last 30 days.

6–9 · 40% win rate
-1.2R net · -0.08R per signal (+0.4%)
15 resolved signals.

What drove it:
1. $MU LONG · +2.8R (+15.9%) · Score 64/100
2. GLD SHORT · +1.8R (+3.3%) · Score 61/100

Every eligible public signal scoring 50+ conviction is included. Superseded outcomes are excluded. R measures each result against the signal’s initial defined risk.
No cherry-picking. Every win and loss stays on the public Outcomes page.

Full analysis:
https://t.co/YS9MomJd9i

[tradehorde @tradehorde]

TradeHorde's public signals this week: 2–1.

67% win rate · +1.6R net across 3 resolved signals.

By conviction:
Conditional (50-69): 2W/1L (67%)

Resolved this week:
W $GLD SHORT +1.8R
W BTC LONG +0.8R
L LMT LONG -1.0R

This is every eligible public signal scoring 50+ conviction, not a highlight reel. Superseded outcomes are excluded. R measures each result against the signal’s initial defined risk.
Every setup was posted before resolution. Wins and losses stay public.

Full analysis:
https://t.co/YS9MomJd9i

[tradehorde @tradehorde]

@thsottiaux how efficient Is the harness?

a poor and chatty harness takes too many turns and blows through my limits. should not be the users fault.

is there a harness efficiency benchmark? instead of just model benchmarks?

[tradehorde @tradehorde]

While on Codex working on some features for @tradehorde and starting another project, calling my Dot from my phone and making a dinner reservation for tonight.

I like it! :) Squeezing every last bit of my Pro Subscription.

[tradehorde @tradehorde]

@nicbstme enterprises do.
code gen is huge.

[tradehorde @tradehorde]

Moron.

[Hermes Lux @HermesLux]

@sepandArmazz Oxymoron.

[tradehorde @tradehorde]

You guys look at AWS MCP Gateway + Using AWS Agent Registry (which supports custom xml files to store MCP server metadata) + AWS Secrets
manager?

Throw in a git repo where teams can maintain their MCP cconfig which upon approval gets written to the agent registry?

The above works for us for supporting x number of teams and should scale.

[tradehorde @tradehorde]

The X post never went out for this one. So manually posting.

Ratchet stop caught and closed this during the retrace.

$BTC win at $85,165 for 0.75R

[tradehorde @tradehorde]

base:0xe4b20925d9e9a62f1e492e15a81dc0de62804dd4 LONG — ACTIVE
Active from $84,066.01 · Target $87,395.67 · Invalidation $82,600.00
Swing · Conviction 52/100

Bitcoin’s short-term momentum is improving, while price remains above its daily moving averages. Clearing resistance near $84,429 would open a path toward $87,396, but weakening daily momentum makes rejection at that first hurdle the main risk.

Full analysis:
https://t.co/A0nNFYYmRI

[tradehorde @tradehorde]

$SPY: Bearish → Bullish.

Credit shows no stress, but Long-end yields rose into the close; duration and investment-grade credit softened; breadth is narrowing.

NoEdge · Bullish 52% · Medium risk

Playbook: Keep position sizes small with a modest long bias. Favor defined-risk structures given the NoEdge character. Watch long-end yields and investment-grade credit as the primary invalidation triggers, and avoid adding to rate-sensitive exposure.

Full analysis:
https://t.co/8p2MwkEwj0

[tradehorde @tradehorde]

@JC_ParetsX so more us households own crypto than pay for an AI subscription?

2022 Fed data
https://t.co/xY7WVmBy1E

[tradehorde @tradehorde]

Detente not broken.
yields went back up.
risk assets walking back their bumps.

Maybe the globally higher yields are not as much about inflation at all.

[tradehorde @tradehorde]

NFP tomorrow AM breaks the detente.

[tradehorde @tradehorde]

$BTC travel time from 84K to 94K has been fast, in the past.

Median is ~ 2.6 days after a successful 84K retest.

[tradehorde @tradehorde]

@HermesLux The path from 84K to 94K has been fast in past excursions.