Home › Famous high-roller gamblers

Most Famous High-Roller Gamblers: The Economics of Being a Whale

In May 1990, a Japanese property investor named Akio Kashiwagi sat down at a private baccarat table in Donald Trump’s Atlantic City casino and started betting $250,000 a hand, 70 hands an hour. The casino had not lured him in to gamble. It had lured him in to keep gambling, because Trump’s people had hired a mathematician, Jess Marcum, who had worked out the one thing that mattered: if Kashiwagi could be kept at the table long enough — past about 75 hours — his chance of walking out a winner fell to roughly 15%. That is the whole business. A casino’s most courted customer is the one it is most mathematically certain to beat, and everything the industry calls generosity — the jets, the villas, the credit — is a retention program priced off that certainty. The whale is not the guest. The whale is the product.

Who are the most famous high-roller gamblers?

They are the players whose documented turnover ran into the hundreds of millions or billions — and whose lifetime position, in almost every case that reached a courtroom or a balance sheet, was deeply negative. The names below are famous because the sums are enormous and because several of them ended in litigation, which is the only reason we have hard figures at all. Read the last column before the amounts: the whale is remembered for one big night; the ledger remembers the rest.

One player on that list is not like the others, and the difference is the whole argument. Kashiwagi, Packer, Watanabe, Kakavas, Ye Gon and al-Zayat were comped — flown in, housed, extended credit. Ivey was barred and sued. The industry’s treatment of the two groups is exactly inverted from what a naive reading of “the casino wants winners to leave” would predict, and the inversion is the tell. To see why, start with the jet.

Why does a casino fly a whale in on its own jet?

Because the flight is cheap next to the loss the whale is expected to book — and the casino prices the trip off the loss it can guarantee, not the one that actually happens. Every comp decision runs through one number the public never sees: theoretical loss, or “theo.” It is not what the player loses. It is what the maths says the player will lose on average, and it is fixed the moment the bet size and the game are known:

Now the jet. An ultra-long-range charter capable of a transpacific run costs roughly $15,000–$20,000 an hour in 2025; a round trip of 24 flight hours lands near $430,000. Set that against a single 12-hour visit from a $100,000-a-hand player, whose theoretical loss for the trip is about $1.0 million. The jet alone eats about 43% of that expected loss; add the villa, the food and the rest and the whole package still lands inside the 30–50% band casinos reinvest for their top whales (16Best analysis; charter rates per Amalfi Jets 2025). The jet is not a gift. It is a line item, and the whale funds it himself in expectation.

How much is a whale worth per hour?

Worth scales linearly with the bet, and it is brutal: a whale’s expected loss per hour is about 0.84 times his average bet on baccarat’s edge. That clean ratio — 70 hands times 1.2% equals 0.84 — is why hosts think in bet sizes, not personalities. Here is the meter running, per hour and per 12-hour trip, with the comp budget it supports at the standard reinvestment rates.

All rows are 16Best analysis assuming 70 baccarat hands per hour and a 1.2% blended house edge (banker 1.06% / player 1.24% per Wizard of Odds). Real whales play far more than 12 hours across a relationship; this is one visit.

The comp side of the ledger is just as mechanical. Casinos do not reinvest evenly — they pay more for the games and players that generate the most reliable theo. Slots earn the machine, so the volume is huge, but reinvestment is thin; a negotiating whale on a high-limit table can pull the rate up toward half.

Read the top of that chart carefully. The most generous comp arrangement on the board still hands the house 60 cents of every dollar the whale is expected to lose. There is no reinvestment tier at which the casino stops winning. Which raises the obvious objection: whales do win sometimes. Kashiwagi was up $9.6 million. So why is the house so calm?

If whales sometimes win, why is the house so sure?

Because a single night is variance and a relationship is expectation, and the casino only ever sells the relationship. Return to Marcum’s calculation, because it is the cleanest statement of the whole model that exists. Kashiwagi’s freeze-out was structured to run until he won or lost a fixed amount — he said $12 million, Trump said $10 million. Marcum’s insight was that resolving a swing that large at $250,000 a hand takes so many hands that the house edge, not luck, decides it. Past 75 hours, the player’s win probability was about 15%. When Kashiwagi was up $9.6 million early in the week, the casino did not panic; it kept the game going, because the maths was on the clock. He left down $10 million over six days.

Put numbers on the grind. At $250,000 a hand and 70 hands an hour, 75 hours is $1.31 billion of handle; on a 1.2% edge that is a theoretical loss of about $15.75 million (16Best analysis). Kashiwagi actually lost $10 million — less than expectation, and he still lost. That gap between his $10m actual and his ~$15.75m theoretical is the entire role of luck in the story: it moved the number in his favour, and he lost anyway. This is why the casino prices comps off theoretical rather than actual loss. Actual loss is noisy and occasionally negative; theoretical loss is a near-certainty the house can bank across many trips. Comp the guaranteed number and you are never wrong for long.

The pattern holds all the way up to the record books. As our page on the biggest casino wins of all time shows, the largest single payouts in casino history come not from skilled whales but from the worst-odds games on the floor — lottery-shaped slot jackpots off machines returning under 90% — and only three of the ten largest documented figures ever reached a player at all. A headline win is a variance event on a negative-expectation bet, which is exactly what a whale’s good night is too.

None of this works, though, unless the whale can keep betting after his cash runs out. That is what credit is for.

How do credit lines and junkets keep the whale playing?

Credit removes the natural stop — running out of chips — and lets theoretical loss accumulate to its full expected value; junkets outsource the credit risk and the recruiting in exchange for a cut of turnover. A marker is an interest-free casino IOU. It exists because a player who has to leave the table to visit an ATM is a player whose theo stops accruing. Extend him a seven- or eight-figure line and the grind runs uninterrupted. The litigation record shows how deliberately this is done:

• Kakavas was flown to Crown Melbourne on the casino’s private jet and given rebates and commissions across 28 visits; he turned over A$1.43 billion and the High Court, in 2013, refused to call the inducements unconscionable (Kakavas v Crown Melbourne Ltd [2013] HCA 25).
• Al-Zayat was, in the Court of Appeal’s own words, allowed to keep gambling on credit for six years rather than being sued early — “so that he could lose millions more pounds to them,” as Lord Justice Sedley put it — before Aspinall’s finally called in a £2 million cheque (Aspinall’s Club Ltd v Al-Zayat, 2007).
• Kashiwagi died owing at least $9 million to US casinos and around $10 million to European ones, having paid Trump only $6 million of the $10 million from the May 1990 game.

In Macau, the credit and recruiting are outsourced to junket operators, who advance chips to VIP players, chase the debts (often across a border where casino markers are unenforceable), and take a commission on rolling-chip turnover — every dollar wagered, not every dollar lost. Since the 2021–22 crackdown that jailed the biggest junket bosses, the model has been capped and shrunk.

Macau VIP share and junket counts per igamingbusiness, Yogonet and reported DICJ figures, 2019–2026. Share figures are of gross gaming revenue and are scope-comparable year to year.

The shrinking VIP share is itself a data point for the spine. The whale segment is being deliberately compressed by regulators precisely because it concentrates the credit risk and the money-laundering exposure — the same features that made whales so profitable made them so dangerous to bank. For the scale of the pool these players swim in, our analysis of how much money casinos make puts global gross gaming revenue near US$712 billion in 2024, about $132 in net losses per adult on Earth. The whale is an extreme point on that same distribution, not a different animal.

Does anyone actually beat the house?

Yes — and the casino’s response to them proves the entire argument, because it does the exact opposite of what it does to a whale. Advantage players win by removing or reversing the house edge rather than riding variance against it. When the edge flips, the comp logic flips with it: a player with positive expectation is not a customer to retain but a liability to eject.

Phil Ivey is the cleanest case. Playing punto banco at Crockfords in 2012, Ivey and partner Cheung Yin Sun used edge sorting — exploiting tiny asymmetries in the card patterns to identify high cards — to win £7.7 million. Crockfords refused to pay; the UK Supreme Court ruled unanimously in 2017 that the method was cheating (Ivey v Genting Casinos [2017] UKSC 67). The same pair used the same technique at the Borgata in Atlantic City for $9.6 million, and a US court ordered them to repay $10.1 million in a 2016 judgment later settled. Ivey generated no comps and no jet. He generated a lawsuit.

What does a whale lose over a lifetime?

Almost exactly the house edge on his turnover — smooth, inevitable, and unrelated to any single night’s drama. This is the reframe that turns anecdote into arithmetic, and it is the same lens our companion page on the biggest gambling losses in history applies to the losing side of the ledger. Divide the loss by the turnover and the whale’s lifetime collapses to a percentage that looks a lot like a house edge.

The two look wildly different, and the difference is not personality — it is games and variance. Kakavas played baccarat, whose edge is about 1.2%; grinding A$1.43 billion through it produced a loss of 1.43% of turnover, almost exactly the edge, because at that volume variance washes out entirely. Watanabe’s 15.4% is far above any single game’s edge, which tells you he played high-edge games badly, chased losses, and ran cold — the very behaviour the alcohol-and-painkillers allegations in his lawsuit describe. But note the direction: variance made Watanabe lose more than the edge and Kakavas lose almost exactly it. In the whole set of famous whales, the tail that beats expectation is the rare one, and it never survives the next trip.

Watanabe is also where the comp economics turn coldest. Harrah’s gave him one of the richest comp deals on record — Rolling Stones tickets, $12,500 a month for airfare, $500,000 in gift-shop credit, and a 15% cash rebate on table losses above $500,000. Apply that rebate to his entire loss and it returns at most about $19 million — 15% of $127 million — leaving the house with at least $108 million, or 85 cents of every dollar he lost; and more than that, because the rebate covered table losses above $500,000 while his heavy slot losses returned nothing (16Best analysis). The single most generous comp term in the modern record — a straight cash rebate on actual losses, not the usual slice of theoretical loss — still could not make the whale a good bet for himself. That is the number to keep. Not the jet. Not the villa. The 85 cents.

Why do whale stories disagree?

Because “how much a high roller lost” is at least five different measurements, and casino PR, junket marketing and court filings each pick the one that flatters them. Here is the disambiguation behind every figure above.

Theoretical versus actual loss. This is the master distinction, and most coverage misses it entirely. Theo is the expected loss the maths guarantees over the volume played; actual loss is what happened after variance. Casinos comp off theo because it is stable; headlines quote actual because it is dramatic. Kashiwagi’s actual $10 million and his theoretical ~$15.75 million over the freeze-out are both true and mean different things — the gap is luck, and it ran in his favour while he still lost.

Single session versus lifetime net. A whale is remembered for one night — Kashiwagi up $9.6 million, Packer’s eight-figure MGM win — but the ledger that matters is the multi-year net, which for every comped whale in the documented record is negative. Ranking a peak session against a lifetime figure, as most listicles do, compares two different quantities.

Turnover (handle) versus loss, and “wagered” versus “drop.” When a court says Kakavas “turned over A$1.43 billion,” that is total amount wagered, hand after hand — not money he brought or lost. Al-Zayat’s “£91.5 million wagered” is more ambiguous: press and pleadings often use “wagered” loosely for the drop (cash converted to chips) rather than true handle, which is why his 25% loss-to-“wagered” ratio is not comparable to Kakavas’s 1.43% of true turnover. We report each figure at the scope its source used and do not divide one basis by another. This is the same handle-versus-revenue error we untangle across the site, applied to individuals.

Credit line versus actual loss. A whale’s reported “debt” is unpaid markers, not his loss. Kashiwagi’s ~$9 million owed at death is money extended and not repaid, a different quantity from what he lost at the table. Casino PR sometimes cites a big credit line as if it were a loss; it is a loan.

Currency and era. Kashiwagi’s 1990 dollars, Kakavas’s Australian dollars and al-Zayat’s 1990s–2000s pounds are not interchangeable, and we have not retro-converted them, because a cross-currency, cross-decade conversion would introduce an error larger than several of the gaps being compared. Where we state a ratio (loss ÷ turnover), both numbers sit in the same currency and the same case, so the ratio is clean even when the raw amounts are not.

Key takeaways

• The whale is the product, not the guest. Comps — jets, villas, credit — are a retention program priced off theoretical loss, the expected number the house can bank, not the actual result of any night.
• The meter is linear and merciless. On baccarat’s ~1.2% edge at 70 hands an hour, a whale’s expected loss per hour is about 0.84× his average bet — roughly $84,000 an hour at $100,000 a hand (16Best analysis).
• No reinvestment tier stops the house winning. Even a negotiated 40% comp rate leaves the casino 60 cents of every expected dollar; Watanabe’s rare 15% cash rebate still left the house at least 85 cents of every dollar he actually lost.
• A night is variance; a relationship is expectation. Kashiwagi was up $9.6 million and left down $10 million because, past 75 hours, his win probability was about 15%. The house sold the relationship and kept the edge.
• Lifetime, the whale loses roughly the house edge. Kakavas lost 1.43% of A$1.43 billion turned over — baccarat’s edge, made real by volume. Variance only ever moves a single trip.
• The advantage player proves the rule. Ivey won by flipping the edge and got sued for $10.1 million and barred. Casinos reward negative expectation and punish positive expectation — the comp is a function of the maths, not the money.

Frequently asked questions

How do casinos decide how much to comp a high roller?

Off theoretical loss, not actual results. A host multiplies the player’s average bet by hands per hour, hours played and the game’s house edge to get expected loss (“theo”), then reinvests a share of it — typically 25–40%, with top whales negotiating 30–50% — as jets, suites, food and credit. A $100,000-a-hand baccarat player generates about $84,000 an hour in theoretical loss, so a full comp package is small against one trip.

Who is the most famous high-roller gambler?

Several compete, by different measures. Akio Kashiwagi is famous for a $250,000-a-hand baccarat freeze-out at Trump Plaza in 1990. Kerry Packer is famous for eight-figure swings and enormous tips across Las Vegas and London. Terrance Watanabe holds the documented modern record for a single year’s losing, wagering over $825 million and losing nearly $127 million in 2007. Harry Kakavas turned over the largest documented sum, A$1.43 billion at Crown Melbourne.

Do high rollers ever actually beat the casino?

On a single night, often — variance cuts both ways. Over a relationship, almost never, because the house edge grinds any large volume of play toward its expected value. Kashiwagi was up $9.6 million mid-week and still left down $10 million. The only players who beat casinos over time are advantage players such as card counters and edge sorters, who reverse the edge rather than gamble against it — and casinos bar them rather than comp them.

What is theoretical loss and why does it matter?

Theoretical loss is a player’s expected loss given how they bet, calculated as average bet times hands per hour times hours times house edge. It matters because casinos price comps off it, not off what a player actually loses. Actual loss is noisy and occasionally negative; theoretical loss is a near-certainty across many trips. Comping the guaranteed number is why the relationship is profitable even on nights the whale wins.

How much did Terrance Watanabe lose, and did his comps help?

Watanabe wagered more than $825 million and lost nearly $127 million at Caesars and the Rio in 2007 — 15.4% of every dollar staked. His comp deal was among the richest on record, including a 15% cash rebate on table losses above $500,000. Applied to his entire loss that rebate returns at most about $19 million, leaving the casino with at least 85 cents of every dollar he lost — more, since his slot losses never qualified. He later sued Harrah’s and settled.

How do junkets and credit lines work for whales?

A credit line, or marker, is an interest-free casino IOU that lets a whale keep betting after his cash runs out, so his theoretical loss can accumulate to its full expected value. In Macau, junket operators advance the chips, chase the debts across borders and take a commission on rolling-chip turnover — fixed at 1.25% since the post-2021 crackdown. Whales such as Kashiwagi and al-Zayat died or were sued owing millions in unpaid markers.

Why are advantage players barred while losing whales are courted?

Because the comp system protects the house edge, not the money on the table. A whale has negative expectation, so the casino profits from his continued play and invests in keeping him. An advantage player — a card counter or edge sorter like Phil Ivey — has positive expectation, so every hour he plays costs the casino, and it ejects him. Ivey won £7.7 million at Crockfords and $9.6 million at the Borgata; both were reversed in court.

Sources

• Wikipedia — Akio Kashiwagi: Trump Plaza freeze-out, Marcum 75-hour calculation, debts at death
• ABC News — Gambler dropped $127M in a year (Terrance Watanabe)
• Above the Law — Watanabe v. Harrah’s: 15% loss rebate, $12,500/month airfare, gift-store credit
• Wikipedia — Kakavas v Crown Melbourne Ltd [2013] HCA 25: A$1.43bn turnover, A$20.5m loss
• South China Morning Post — Australian high-roller Harry Kakavas loses lawsuit against casino
• Court of Appeal (England & Wales) — Aspinall’s Club Ltd v Al-Zayat: £91.5m wagered, £23.2m lost, £2m marker
• Casino.org — Kerry Packer: MGM wins, $16.5m (A$28m) 1999 London loss, tipping
• Al Jazeera — Australian tycoon Kerry Packer dies (2005)
• US Department of Justice — Las Vegas Sands returns $47.4m over Zhenli Ye Gon transfers
• Card Player — Phil Ivey settles $10.1 million Borgata edge-sorting lawsuit
• Wikipedia — Ivey v Genting Casinos [2017] UKSC 67 (Crockfords £7.7m)
• Wizard of Odds — Baccarat house edge: banker 1.06%, player 1.24%
• TravelZork — Casino theo / Average Daily Theoretical: the comp formula
• Duetto — Casino comp reinvestment strategy (share of theoretical loss)
• Amalfi Jets — Private jet charter cost per hour by category, 2025–26
• iGaming Business — Macau junket cap, licensing and 1.25% rolling-chip commission
• Yogonet — Macau approves 29 casino junkets for 2026