Thursday, March 12, 2026

Hippos Handicapping - LLM Virtual Panel - Pertemps Network Final Handicap Hurdle Review

WCMI Hippos Handicapping - LLM Virtual Panel - Pertemps Network Final Handicap Hurdle Review

The Hippos Handicapping Panel — where memory and mechanisms collide, but only the horses decide.

Our ongoing exploration of the role of Large Language Models (LLM) in sports trading.

Welcome to the Hippos Handicapping Panel — a virtual round‑table of racing minds brought to life with the help of an LLM. Each Hippo has a distinct voice:

  1. Mick – Aussie handicapper and professional punter
  2. Pearl – Canadian academic and causal analyst
  3. Philip – British host who keeps them honest and sneaks in his own Weekend Warrior longshots

Together they blend events and explanations into a lively debate that is equal parts analysis and paralysis.

Hippos Handicapping - LLM Virtual Panel - Pertemps Network Final Handicap Hurdle Review

🐴 Hippos Handicapping — LLM Virtual Panel — Post-Race Review

Cheltenham | Thursday 12 March 2026 | Pertemps Network Final Handicap Hurdle | 2m 7f 213y | 24 runners

🎙️ Philip Opens

Philip: Well. Good evening, and welcome to what I can only describe as the most comprehensive humbling this panel has experienced since we started doing these reviews. Let me state the facts plainly and without adornment, because the facts are damning enough on their own. The Pertemps Network Final Handicap Hurdle was won by Supremely West, returned at 100/30, trained by Dan Skelton, ridden by Harry Skelton. He travelled strongly throughout, made headway to join the leaders before five out, took it up before the last, and kept on well to win by three and a half lengths from Lavida Adiva at 22/1. Third was Ikarak at 40/1, beaten six lengths. Fourth was Letos at 22/1.

Now, for those of you who didn't read our preview — and frankly, if you followed our advice, I envy your ignorance — let me remind the audience of the panel's position heading into this race. We identified Supremely West, unanimously, as the "false favourite." Those were our words. We labelled him a horse who had "no business being favourite for a Cheltenham Festival handicap." We built an entire analytical edifice around opposing him. Our consensus win selection was Kikijo at 16/1. Kikijo finished eighteenth. Our structural fit selection was Electric Mason at 11/1. Electric Mason finished tenth. Our consistent form selection was Ace Of Spades at 12/1. Ace Of Spades finished thirteenth. My Weekend Warrior, Staffordshire Knot, pulled up before the last. This is, and I'm choosing my words carefully here, a total and unmitigated disaster. Mick, I think you owe the audience an explanation. Possibly an apology. Where do we even begin?

🗂️ Mick — Post-Race Reflection

Mick: Yeah, look. No dressing this up. No wriggling out of it. No "approximately right" about this one, mate — we were precisely, spectacularly, catastrophically wrong. I sat here yesterday and told the audience that Kikijo beating Supremely West by a combined fifty-six lengths in two runs was the story of the race. I said the market was "insane" to have Supremely West as favourite. I said — and this is the bit that's going to haunt me — "the form is the form." Well, the form wasn't the form. Or more accurately, I was reading the wrong form.

Let me face the music, horse by horse. Kikijo was my primary win selection at 16/1. He went off 14/1, so the market actually agreed with me and shortened him. And he finished eighteenth. Eighteenth out of twenty-four. The race comment says he was prominent, in touch with the leaders at halfway, and then "weakened on the turn before last." He didn't just fail to win — he collapsed. He had nothing left when the race got serious. And the reason is staring me in the face: the ground. Those two demolition jobs over Supremely West came on Soft at Cheltenham and Soft at Sandown. Thursday's ground was Good, and Kikijo simply couldn't reproduce that form on a faster surface. Pearl flagged this in the preview. She said the margins might compress on Good ground but she didn't think the causal mechanism would reverse entirely. She was half right — it didn't just compress, it inverted. Kikijo was a different horse on Good ground, and I didn't weight that variable heavily enough. Lesson learned the hard way.

Ace Of Spades, my each-way play at 12/1, went off at 16/1 — actually drifted from the preview price, which should have been a warning signal — and finished thirteenth. The race comment is telling: "prominent, in touch with leaders before fourth, not fluent three out, weakened before two out." He was never travelling and the mistake at three out killed whatever chance he had. To be fair to the horse, he's from the Skelton yard that actually produced the winner, but I picked the wrong Skelton runner. That stings.

Champagne Chic at 20/1, my value swing, went off 18/1 and finished twelfth. "Held up in rear, some headway before last, stumbled after last, no impression." He was never competitive. The progressive lightweight theory was nice in the abstract but didn't translate to the reality of a Festival handicap where he simply wasn't good enough.

And then there's the winner. The horse I spent four hundred words telling everyone to oppose. Supremely West travelled strongly — the race comment couldn't be clearer about that — headway to be in touch with leaders before five out, led before the last, kept on well. This wasn't a scrappy, lucky handicap win. He dominated the finish. He looked like, dare I say it, the best horse in the race. On Good ground, with a prominent position, he was a completely different proposition to the horse who got hammered on Soft ground at Cheltenham and Sandown.

The biggest mistake I made was treating form on Soft ground as transferable to Good ground without proper adjustment. I've seen it a hundred times when other punters do it, and I've always said "different surface, different race." But when Kikijo's margins were so enormous — twenty-four and thirty-two lengths — I convinced myself the gap was too big for conditions to close. It wasn't. The ground didn't just close the gap, it reversed the order entirely. Seen this before, actually, and I should have known better. A good smack around the head is sometimes the best education money can buy.

🎙️ Philip to Pearl

Philip: Pearl, in the preview you were the one who actually raised the ground question head-on. You said, and I'm quoting, "I don't think the underlying causal mechanism reverses entirely." But it did reverse — completely. Your primary selection Electric Mason finished tenth, your each-way Kikijo finished eighteenth, and your progressive risk Absolutely Doyen finished eighth. The winner was the horse all three of us opposed. So my question to you is surgical: your causal framework identified the right variable — the ground — but then drew the wrong conclusion from it. What went wrong in the reasoning chain?

🔗 Pearl — Causal Post-Mortem

Pearl: This is a genuinely instructive failure, and I want to be precise about where the reasoning broke down, because the breakdown is more interesting than the failure itself.

My framework correctly identified ground as the key mediator in the Kikijo-Supremely West dynamic. What I got wrong was the magnitude of the effect. I treated ground as a continuous variable — one that would compress margins proportionally — when in reality it operated as something closer to a threshold variable. Below a certain firmness, Kikijo's stamina and grinding style dominated. Above that threshold, on genuine Good ground, his advantage evaporated entirely and Supremely West's latent class reasserted itself. The relationship wasn't linear. It was binary, or close to it. That's a fundamental modelling error on my part, and it contaminated every downstream prediction I made.

Let me take my selections in order. Electric Mason at 11/1 was my primary pick, and he went off at 9/1 — the market shortened him significantly, which means I wasn't alone in my assessment. He finished tenth. The race comment reads "soon prominent, lost position before second, headway and in touch with leaders before fourth, dropped to midfield before seventh, no impression." That's a horse who was struggling from the early stages, never found a rhythm, and couldn't sustain any forward momentum. I flagged the four-month layoff as a risk factor in the preview but argued it was a "planned absence" rather than a symptom. The evidence suggests I was wrong about that, or at least that even a planned absence left him short of peak fitness for a Festival cavalry charge. The confounder won.

Kikijo at 16/1 as my each-way pick — I've already agreed with Mick's diagnosis. The ground mediated the entire form picture and I underestimated its effect. I said in the preview that his Aintree fourth on Good ground, beaten three lengths, proved he "handled a drier surface adequately." But handling a surface adequately and handling it well enough to win a Festival handicap are categorically different things. I committed a classic error of conflating sufficiency with optimality.

Absolutely Doyen at 14/1 as my progressive risk selection finished eighth, off a starting price of 8/1 — he was heavily backed. The race comment says he was "in touch with leaders, midfield before fourth, outpaced after two out, rallied before last, not reach leaders." That's actually a more creditable run than I feared. He wasn't disgraced — he ran his race — but the five-timer was always against the base rates for horses stepping up to Festival level, and so it proved. Interestingly, my caveat in the preview was that his opposition quality in those five wins was a confounding variable, and the Festival exposed exactly that. A horse whose best RPR came at Musselburgh could not sustain that level at Cheltenham. The confounding variable I identified was the right one. I just should have weighted it more heavily and moved him down the pecking order rather than including him at all.

Now, the critical failure. I need to address why I opposed Supremely West. My causal framework said: this horse has been beaten a combined fifty-six lengths by Kikijo in two runs, therefore he is inferior. But I failed to condition on the right variable. Those defeats were on Soft ground. On Good ground, the entire causal chain that produced those defeats ceases to operate. I should have asked: what is Supremely West's form on Good ground specifically? His earlier form, before those two hammerings, included a third at this course on better ground. His official rating of 135 was earned through performances that the handicapper believed warranted that mark. And crucially, as a Dan Skelton-trained eight-year-old with Harry Skelton in the saddle, he represented a stable that knows how to target Festival handicaps. The data spoke, but I was listening to the wrong frequency.

🎙️ Philip Challenges Both

Philip: Let me be blunt with both of you, because the audience deserves it. You each identified pieces of the puzzle — Mick saw the collateral form, Pearl saw the ground variable — but you assembled them into the wrong picture. The winner was staring at you from the top of the market at 7/2, and your combined response was to label him a fraud. But here's what really troubles me: the placed horses. Lavida Adiva at 22/1 finished second. Ikarak at 40/1 finished third. Letos at 22/1 finished fourth. None of these horses received a single mention in our preview. Not one word. We discussed nine different selections across the panel — Kikijo, Electric Mason, Ace Of Spades, Champagne Chic, Absolutely Doyen, Staffordshire Knot, and we banged on about Supremely West for half the preview — and we didn't even nod toward any of the placed horses. Mick, is the honest answer simply that in a twenty-four-runner handicap, we overthought this so badly that we talked ourselves past the obvious?

🗂️ Mick — Rebuttal

Mick: Yeah, Philip, and it's worse than overthinking. It's a case of narrative capture. We built a story — "Supremely West is a false favourite, Kikijo is the key" — and every piece of analysis we did after that was in service of that story. Pearl's ground analysis was used to test whether Kikijo would still beat Supremely West, not whether Supremely West might actually win. My collateral form work was one-dimensional, as you pointed out in the preview itself. You literally asked me, "aren't you making a one-dimensional case?" and I batted it away by pointing to Kikijo's RPR and the handicapper's mark. You were right and I was wrong.

On the placed horses — Lavida Adiva was trained by Ruth Jefferson and ridden by Brian Hughes, a northern combination that wouldn't have been top of anyone's Festival shortlist, but the horse ran a stormer, held up and making steady headway, went second before the last. Ikarak at 40/1 for Olly Murphy and Sean Bowen rallied on the run-in from the rear. Letos at 22/1 was the Anthony Mullins and Danny Mullins combination from Ireland — travelled strongly, prominent throughout, just found the hill too much in the final hundred and ten yards. Could I have found any of these? Honestly, probably not on the evidence I had. But that's the point about twenty-four-runner handicaps. Sometimes the honest answer is "I don't know," and the bet is smaller or doesn't happen at all. Instead, I went in heavy on Kikijo as a "consensus win" and the result was an eighteenth-place finish. That's the price of conviction without humility.

The one thing I'll say in partial defence is that the market as a whole didn't find these horses either. Lavida Adiva went off at 22/1, Ikarak at 40/1, Letos at 22/1. This wasn't a case where the market screamed "look over here" and we ignored it. The entire betting public was largely blind to the placed horses. The only entity that wasn't blind was the favourite's connections, and they turned out to be right.

🔗 Pearl — Rebuttal

Pearl: I want to add a structural observation to Mick's honest assessment. When I look at the first four home — Supremely West, Lavida Adiva, Ikarak, Letos — they share a common tactical profile. The race comments tell the story: Supremely West was "held up in midfield" before making headway from five out. Lavida Adiva was "held up in midfield" before making steady headway from two out. Ikarak was "held up in rear" before steady headway from two out. Letos was "prominent" and "travelled strongly" throughout. Three of the first four were held up and came from behind, making their moves late, while the one who raced prominently had the benefit of travelling strongly throughout on ground that suited.

Now compare that with our selections. Kikijo was "prominent, in touch with leaders halfway, weakened on turn before last" — he raced too prominently on ground that didn't suit and emptied. Electric Mason "soon prominent, lost position" — he was in the wrong position from the start. Ace Of Spades was "prominent, not fluent three out, weakened" — again, too forward too soon. Our entire selection set was populated by horses that either raced prominently and couldn't sustain it, or needed different ground to produce their best. The New Course configuration I discussed in the preview — with its longer run-in favouring horses who can grind up the hill on a protracted approach — actually played out exactly as described. We just backed the wrong horses for that profile.

The counterfactual I should have explored is: what happens to Supremely West when you remove Soft ground from the equation? Instead, I asked: what happens to Kikijo when you remove Soft ground? The asymmetry in my questioning was the asymmetry in my error. I investigated the risk to my preferred selection rather than investigating the upside of the horse I was opposing. That's confirmation bias dressed up in causal language, and I should know better.

🎙️ Philip's Synthesis

Philip: Let me try to salvage some educational value from the wreckage. The panel's fundamental error was not analytical — it was epistemological. We confused a specific, ground-dependent form line with a universal truth. Kikijo beat Supremely West by large margins on Soft ground, and we promoted that observation to a law. But it wasn't a law. It was a conditional statement that applied on Soft ground and failed on Good ground. The moment conditions changed, the conclusion evaporated.

There's a deeper lesson about the architecture of our reasoning. Mick built his case from collateral form — his case-based reasoning drew on vivid, memorable margins. Twenty-four lengths and thirty-two lengths are dramatic numbers that lodge in the mind. Pearl's causal framework correctly identified ground as a mediating variable but drew the wrong conclusion about its magnitude. And I, as the supposed synthesist, saw the two analytical streams converge on the same answer and mistook convergence for validation. Three people agreeing on the wrong answer doesn't make it right. That's the textbook definition of groupthink, and I flagged it in my own summary — I said our consensus was "either a sign of genuine insight or a spectacular case of groupthink." Turns out it was the latter.

What would have pointed us toward the winner? Honestly, the market itself. Supremely West was 7/2 favourite for a reason. The Skelton yard had targeted this race specifically. Harry Skelton chose to ride him over Ace Of Spades, which tells you which horse the yard fancied. The official rating of 135 was competitive. And his earlier Cheltenham form — a third at the course — showed he could handle the track. We dismissed all of this as "anchoring on a name and the Skelton brand." Sometimes the crowd is right, and the clever money is stupid.

On the positive side — and there isn't much — the panel's observation about the New Course configuration proved entirely correct. Held-up horses who could grind up the long run-in dominated the finish. We just failed to identify which held-up horses would do the grinding. And Pearl's caution about Absolutely Doyen's five-timer was vindicated — he finished eighth, unable to sustain his unbeaten record at Festival level. But these are crumbs from a very sparse table.

🧢 Weekend Warrior Review

Philip: And so to the Weekend Warrior, which this week reads less like a speculative flutter and more like a cautionary tale. My pick was Staffordshire Knot at 20/1, the top-weighted Grade 2 winner with Jack Kennedy aboard. I made a stirring case about class prevailing, about Kennedy choosing this horse out of the entire Elliott string, about twelve stone not being an insurmountable burden in a compressed handicap. I even said, with characteristic modesty, that if he landed a blow I'd be insufferable until the Gold Cup.

Well. Staffordshire Knot was pulled up before the last. The race comment is grim: "jumped right on occasions, prominent, lost position before three out, weakening when jumped badly right two out, pulled up before last." Kennedy reported the gelding jumped badly right-handed throughout, and the vet had nothing to report. So this wasn't a fitness issue or a ground issue — the horse simply wouldn't jump straight on the track. He went off at 14/1, shortened significantly from the 20/1 I quoted, so real money came for him, which means I wasn't alone in my delusion. But a pull-up is a pull-up, and no amount of narrative dressing can disguise a horse that refused to co-operate.

The lesson is one I should have tattooed on my forehead by now: top weight in a twenty-four-runner Festival handicap is a structural headwind, not a style choice. The class argument sounded wonderful in the preview, and the connections clearly believed in the horse given the market move, but when a horse is carrying twelve stone in a staying handicap and then adds jumping issues on top, you're fighting gravity. I will not be insufferable until the Gold Cup. I will not be insufferable until the weekend. I will be buying the first round at the bar and hoping everyone forgets by morning.

📋 Key Takeaways

  • Ground is not a modifier — it can be a transformer. The panel treated the shift from Soft to Good as a variable that would compress margins, when in fact it fundamentally rewrote the form hierarchy. In future Pertemps reviews, collateral form from different ground conditions must be treated with far greater scepticism, especially when the margins are extreme. Extreme margins on one surface may reflect surface-specific superiority, not absolute class.

  • Beware of narrative capture in consensus picks. When all three panelists converge on the same opposition play — in this case, opposing Supremely West — the convergence itself should trigger a stress test. The panel should ask: "Are we converging because we've independently found the same evidence, or because we're building on each other's assumptions?" Here, Mick's collateral form case fed Pearl's causal framework, which fed Philip's synthesis, creating a closed loop of mutually reinforcing error.

  • The market favourite deserves the presumption of innocence. At a meeting as fiercely competitive as Cheltenham, the favourite is favourite for a reason. The crowd isn't always right, but opposing a 7/2 shot requires more than one line of form — it requires evidence that the horse cannot win under the conditions of the actual race, not just the conditions of its worst runs.

  • Held-up horses thrived on the New Course configuration. The panel's structural analysis of the Thursday course — longer run-in, later hurdle positioning, premium on stamina over tactical speed — was correct. Three of the first four were held up and produced late surges. This is a replicable insight for future New Course handicaps.

  • Layoff horses struggled. Electric Mason (four months off, finished tenth) and the panel's dismissal of freshness as a factor proved costly. Campaign-hardened horses who had been through the qualifying route had the fitness edge in a genuinely run staying handicap.

  • Jockey choice is a signal, not noise. Harry Skelton chose Supremely West over Ace Of Spades from the same yard. Jack Kennedy chose Staffordshire Knot but the horse let him down through jumping. When a top jockey picks one stable companion over another, that's information about the yard's private view, and we underweighted it for Supremely West while overweighting it for Staffordshire Knot.

💭 Final Thought

Philip: The philosopher Karl Popper once said that the mark of a genuine theory is its capacity to be proved wrong. Well, our theory about this race — that Supremely West was a false favourite, that Kikijo was the answer, that the collateral form was dispositive — was proved wrong in the most emphatic fashion imaginable. The horse we told you to oppose won by three and a half lengths. The horse we told you to back finished eighteenth. The horse I picked as my longshot didn't finish at all. In the cold light of a Cheltenham evening, that's the game. You study, you reason, you construct your arguments with all the rigour and wit you can muster, and sometimes the horse just runs fast and you look like an idiot. The only honest response is to learn, recalibrate, and come back tomorrow with a better question. Which is precisely what we intend to do. Good evening.

Generated by Hippos Handicapping — LLM Virtual Panel [ https://vendire-ludorum.blogspot.com/ ]

Posted by at
Labels: , , ,