HARTIGAN, Circuit Judge.
This is an appeal from a judgment of the United States District Court for the District of Rhode Island entered on May 20, 1955 which sentenced the defendant to imprisonment for a term of one year and to pay a fine of $5,000 after verdict of guilty by a jury.
The information filed by the United States Attorney charges “That during the calendar year 1951 George Winkler * * * had and received a gross income in excess of $600; that by reason of such income he was required by law, [768]*768after the close of the calendar year 1951, and on or before March 15, 1952, to make an income tax-return to the Collector of Internal Revenue for the District of Rhode Island stating specifically the items of his gross income and any deductions and credits to which he was entitled, that well knowing all of the foregoing facts, he did wilfully and knowingly fail to make said income tax-return to the said Collector of Internal Revenue * * * in violation of Section 145(a), Internal Revenue Code of 1939; 26 U.S.C. § 145(a).”
The Government filed on March 8, 1955 a bill of particulars which in substance sets forth that the defendant’s income during the calendar year 1951 had been received from bookmaking (gambling) activities, “that the amount of total gross income was $22,328 received from bookmaking and that the total gross income, as set out above, represents gross receipts and is determined without reduction for losses.”
Under § 22 of the Code, 26 U.S.C.A'. § 22, gross income is defined to include “ * * * gains or profits and income derived from any source whatever.” We are concerned here with the question of whether a professional gambler has a gain or profit within the statutory meaning every time he wins a bet with one of his customers. If the answer to this question is in the negative, the judgment below must be reversed.
The evidence introduced by the Government tended to establish the following facts: that the defendant’s income was from gambling; that he had never filed any- income tax returns; that there was no record of the defendant’s filing an income tax return for 1951; that he had been a gambler all his life; that over the years, and with no specific reference to 1951, he had made large amounts of money in the booking and gambling business although the precise amount was not specified.
Vincent Sorrentino, a manufacturing jeweler who was the chief witness for the Government, testified that he knew Winkler for about fifteen or sixteen years and placed bets with him for ten or eleven years; that Winkler gave Sor-rentino a $1,000 credit to start and that later this was increased to $2,000; that Winkler on Sunday or Monday used to mail Sorrentino slips showing the daily bets Sorrentino placed with Winkler; that the records and slips which Sorren-tino would receive showed the bets and also the minus and plus if Sorrentino owed money or Winkler owed Sorrentino money.
The Government also introduced in evidence a cheek of the Uneas Manufacturing Company, signed by Vincent Sor-rentino, Treas., dated September 8, 1951, payable to Winkler in the sum of $1280 which Sorrentino testified was “in payment of money I owed to George Wink-ler for bets placed with him.”
Sorrentino admitted that Winkler paid him some money in 1951 but he could not recall the amount although he admitted it was over $1,000.
A number of these weekly slips for 1951 were introduced into evidence by the Government but they admittedly did not cover all the betting transactions between Sorrentino and Winkler for the entire calendar year of 1951.
John M. Mullen, Special Agent of the Intelligence Division of the Internal Revenue Service, testified that he made an analysis and recapitulation of these slips; that the amount bet was $22,238; that the winnings on bets placed by Sor-rentino with Winkler were $17,364; that there was a gain to Winkler of $4,964. Mullen further testified that in making this total recapitulation he did not take into consideration any payments that might have been paid by Winkler to Sor-rentino and he did not try to ascertain how much money Sorrentino received from Winkler. He testified “I just computed these figures on these slips.”
At the conclusion of the Government’s evidence the defendant, pursuant to Rule 29, Fed.Rules Crim.Proc. 18 U.S.C.A., [769]*769filed a motion for judgment of acquittal which was denied. The defendant then rested his case.
In his charge to the jury the trial judge said:
“As I stated to you a moment ago, there are two issues in this case: First, whether or not the defendant received a gross income in excess of $600 during the calendar year 1951; and, secondly, whether or not, having received that gross income in excess of $600, he did knowingly and willfully fail to file a tax return.
“With respect to the first of these issues, I instruct you that it is sufficient if the Government has established by proof beyond a reasonable doubt that the defendant during the calendar year 1951 had and received winnings in his occupation as a bookmaker in excess of $600.
“Under the Internal Revenue Code the term, ‘gross income’, includes gains, profits, and income derived from compensation for personal services of whatever kind and whatever form paid. It also includes gains, profits, and income derived from interest, dividends, securities, gains or profits or income derived from any source whatever. It makes no difference, as a matter of law, whether such income was derived from lawful or unlawful activities.
“Under the provisions of the Internal Revenue Code a person engaged in the business of bookmaking must, if the bets won by him during the calendar year exceeded the sum of $600, file a tax return showing the total amount of bets won by him. Under the Code he is then entitled, in order to arrive at his income tax, to deduct the total amount of bets lost by him up to but not in excess of the amount of bets, of the value of bets won by him.
* * * •» -x- *
“ * * * n js not necessary for the Government to establish the precise amount of gross income received by the defendant or the precise amount of the bets or wagers won by him. It is sufficient if the Government has established by proof beyond a reasonable doubt that during the calendar year 1951 the defendant had and received income from winning bets in excess of $600.”
Among other assignments of error, the defendant contends that the trial court committed prejudicial error in making the above charge to the jury concerning the meaning of gross income as the term relates to the defendant.
It is entirely clear that the Government’s evidence will support a conviction only upon one of two theories of gross income: (a) total receipts, or (b) total winning bets. The trial court rejected the former theory and submitted the case to the jury upon the latter.
Before turning to a consideration of the cases which we cite below, it might be helpful to analyze in some detail the nature of the professional gambler’s operation. In a certain sense a bookmaker may be likened to a merchant, one major difference being that instead of selling articles at a price exceeding the cost, the bookmaker, in effect, utilizes the laws of probability in such a way as to enable him to sell a certain indeterminate amount of money for a larger indeterminate amount of money, over an extended period of operation.
Let us consider the following somewhat over-simplified illustration. The operator of a lottery makes the following offer to each of ten prospective customers: “I will sell you one ticket for $1,-000, and I promise to buy back or redeem one of these ten tickets at a price of $9,000. The winning $9,000 ticket will be selected by lot.”
If we consider this transaction as in the nature of a sale of $9,000 for $10,000, then it would seem apparent that the $9,000 is actually “cost of goods [770]*770sold” and as such necessarily excluded from income.1
Regulations 118, Income Tax Regulations (1953);
“Sec. 39.22(a)-5 Gross income from business
■ “In the case of a manufacturing, merchandising, or mining business, ‘gross income’ means the total sales, less the cost of goods sold, plus any income from investments and from incidental or outside operations or sources. In determining gross income, subtractions should not be made for depletion allowances based on discovery value or percentage of income, selling expenses, or losses, or for other items not ordinarily used in computing cost of goods sold. But see § 39.23(m)-l(e).”
Moreover, the justification for excluding returns of capital from gross income goes deeper than mere Treasury Regulations and rests ultimately upon constitutional concepts. There is general agreement “ * * * that amounts received as a return of capital or investment are not income within the general meaning of that term and may not be taxed under the Sixteenth Amendment.” Mertens Law of Federal Income Taxation, Yol. 1, § 5.06 (1942).
Approaching the problem from another direction, it would not seem proper to regard the $9,000 as merely a deductible loss since in the consideration of the deductibility of losses “ * * * it must be remembered that the statutory deductions are a matter of legislative grace. Questions of constitutionality do not intrude into the discussion; only those losses which are specifically allowed by statute may be deducted.” Paul and Mertens Law of Federal Income Taxation, Vol. 3, § 26.25 (1934). Of course it might be said that the operator of the $9,000 lottery is not selling $9,000 in actual cash, but is only selling his promise to pay $9,000 in cash. But even in this view it might be argued that a ticket which represents a reliable promise to pay $9,000 is in itself a capital asset.
Assuming, however, that the sale .of these tickets did not constitute the sale of a capital asset, it nonetheless seems perfectly clear that the $10,000 was received only because a promise was made to pay out $9,000. And when it came into the possession of the lottery operator it was already impressed with and subject to a fixed obligation to pay out all but $1,000 of the total receipts. The mere fact that this obligation might not in some instances be legally enforceable seems of no consequence since from a business point of view the operator must clearly pay or cease operations regardless of the legal status of his debts. Under these circumstances we conclude that the operator of the lottery has a “gain” of only $1,000 within the meaning of § 22 of the Internal Revenue Code because only $1,000 came to him from the operation.
This view of the transaction is further supported by the basic nature of the loss deduction under § 23 of the Code, 26 U.S.C.A. § 23, which historically contemplated, among other losses, those resulting from fire, storm, shipwreck, and theft — in short, various unintentional misfortunes which every taxpayer naturally hopes to avoid. Obviously, it seems to us, the lottery operator’s $9,000 payment is of an entirely different nature, being intentionally incurred and representing the necessary inducement which he must offer his customers in order to make his business productive. Accordingly, we believe that [771]*771this $9,000 must he regarded as exclusion from gross income rather than a “loss” under § 23 which Congress could in its discretion deny the taxpayer.
Of course the operation of a bookmaker differs from a lottery in some respects, chief among which is the possibility that losses on any given transaction may exceed gains — that is to say, losing bets may exceed winning bets with respect to any particular race.
This raises additional questions as to what constitutes gross income of a bookmaker as contrasted with the gross income of the lottery operator who knows that he cannot have a net loss on the entire transaction.
There are at least four possible theories upon which a bookmaker might calculate his gross income:
I. (a) Total receipts less total payments to winning bettors over the year.
(b) Total winnings on winning bets less total losses on losing bets over the year.
(e) Total net winnings on winning races less total net losses on losing races over the year.
(These three methods are merely different ways of phrasing the same economic concept leading to the same mathematical result.)
II. Total winnings on winning races over the year.
III. Total winning bets over the year. (This is the theory of the trial court.)
IV. Total receipts over the year.
(This was urged by the Government along with III above, but was rejected by the trial court.)
The practical differences between the four theories set forth above can perhaps best be illustrated by a concrete example. Let us assume that a bookmaker accepts one $200 wager on each horse in each race throughout the year. For purposes of convenience this hypothetical bookmaker’s business covers only twelve races in the course of the year.
Total
Receipts
or Profit
Races Bets Paid Out Winning Bets Losing Bets or Loss
1* $1400 $ 800 $1200 $ 600 +$ 600
2. 1200 600 1000 400 + 600
3. 1200 2000 1000 1800 — 800
4. 1000 400 800 200 + 600
5. 1200 1000 1000 800 + 200
6. 1000 1600 800 1400 — 600
7. 1200 1600 1000 1400 — 400
8. 1200 1200 1000 1000 —
9. 1000 2000 800 1800 — 1000
10. 1000 1800 800 1600 — 800
11. 1200 800 1000 600 + 400
12. 1200 1000 1000 800 + 200
Totals:
$13,800 $14,800 $11,400 $12,400 —$1,000
[772]*772Total profit on winning races — $2,600
Total loss on losing races — $3,600.
Gross income would be calculated as follows:
I. (a) Total receipts less total payments to winning bettors over the year; $13,800 less $14,800, or minus $1,000.
Í. (b) Total winnings on winning bets less total losses on losing bets over the year; $11,400 less $12,400, or minus $1,000.
I. (c) Total net winnings on winning races less total net losses on losing races over the year; $2,600 less $3,600, or minus $1,000.
It will be seen that under this conception gross income would be zero with a net loss from the entire year’s operations of $1,000 which might or might not be deductible under the provisions of § 23. This $1,000 loss is clearly of a kind which Congress could in its discretion properly deny. If it were denied the result would merely be that the taxpayer could not apply his net gambling losses against unrelated income or against income for other years. This would seem to be entirely within the scope of Congressional authority.
II. Total winnings on winning races over the year, or $2,600. This would leave $3,600 lost on losing races as a deduction which Congress could in its discretion deny. This $3,600 breaks down into two logically distinct categories:
A. $1,000 net loss, as indicated above.
B. $2,600 which represents the amount lost on losing races up to the amount won on winning races.
There seems to us a serious question whether this $2,600 may be regarded as a deduction which Congress could in its discretion deny, or whether it ought not properly to be regarded as an exclusion from gross income.
It should be noted that these are not the kinds of payments previously discussed which are unavoidable and deliberately incurred for purposes of inducing customers to make bets. On the contrary, the bookmaker need not ever sustain a losing race — though losing bets, of course, are inevitable. But he knows that losing races are a distinct possibility, perhaps even a probability, on occasion, and he hopes to utilize the laws of probability in such a way that over an extended period his winnings on winning races will exceed these losses on losing races with a net profit resulting over the entire operation.
It seems to us highly doubtful that Congress can, in its discretion, view the business in such a radically different light from that in which the taxpayer himself regards it, with the result that whereas the taxpayer who entered into these transactions with the expectation of balancing losses against gains over the year, and who would be foolish to enter into the business on any other basis, must nonetheless calculate his gross income on the basis of gains on winning races alone, leaving all losses on losing races to be granted as a deduction which Congress may or may not in its wisdom see fit to grant.
In every tax year the bookmaker will have some winning races and some losing races. In the above illustrative case the bookmaker won $2,600 on winning races and lost $3,600 on losing races. Yet, in this view, Congress can tax him on $2,600 despite the fact that his business, as he understands it, and as any businessman would understand it, shows an actual loss of $1,000 over the anticipated yearly period of operations.
III. Total winning bets over the year, or $11,400. This would leave $12,-400 in the category of losses deductible in the discretion of Congress. This figure includes:
A. $1,000 net loss as indicated above.
B. $2,600 loss on losing races up to the amount of winnings on winning races, as above.
[773]*773C. $8,800 which represents the yearly total of the amount of losing bets in each race up to the amount of winning bets in each corresponding race.
Although it is perhaps possible that Congress could in its discretion deny the $2,600 total losses on losing races (the amount by which losing bets exceed winnings bets on any given race) it seems to us completely inconceivable that Congress could, if it chose, deny this $8,800 total which represents losing bets up to the amount of winning bets throughout the year. This figure includes the necessary and inevitable payments in each individual race which provide the inducement to customers, sustaining the entire operation and without which there could not possibly be any winning bets or gains of any kind, taxable under any theory of gross income.
IV. Total receipts over the year, or $13,800. This would leave $14,-800 as a deduction allowable in the discretion of Congress, which breaks down as follows:
A. $1,000 net loss as above.
B. $2,600 total losses in losing races, as above.
C. $8,800 total losing bets up to the amount of winning bets, as above.
D. $2,400 which represents money left with the bookmaker (customer’s $200 winning bet in each of 12 races) which he had only temporarily in his custody and which was returned to the winning bettor in each race along with his winnings.
Clearly this $2,400 is an exclusion from gross income and the trial judge was correct in rejecting the gross receipts theory of gross income which would treat this $2,400 as a loss deductible in the discretion of Congress. It was never the bookmaker’s in any sense and could not constitute a part of his gross income.
We now turn to significant cases. In James P. McKenna, 1925, 1 B.T.A. 326, gross income of a professional gambler was determined to be as set forth in the first of the four theories above — that is, total receipts less total payments over the year, or, in terms of our illustration, $13,800 minus $14,800. In this view the only portion of the total payments for the year to be treated as deductible only in the discretion of Congress is the $1,-000 net loss over the entire yearly operation. All other payments are treated as offsetting gains prior to the determination of gross income.
This case was followed shortly by the case of Mitchell M. Frey, Jr., Ex’rs, 1925, 1 B.T.A. 338, which applied the same concept of gross income to a casual gambler. The Board of Tax Appeals said at pp. 341 and 342, of 1 B.T.A.:
“It now remains for us to consider the amount of the losses sustained by the deceased from his gaming. In his returns he asserted that he had won $900 during the year 1919 and $26,588 during 1920. As a matter of fact he won nothing and his reported winnings should be disregarded in computing the tax. The Commissioner has argued that each gaming venture is a separate item; that those in which winnings were made stand apart from those in which money was lost and that the totals of the winnings must be returned and taxed, while the totals of the losses may not be deducted.
“The fallacy of this contention is well pointed out in Appeal of James P. McKenna, 1 B.T.A. 326, and requires no further comment here. The decedent’s operations extend through annual periods and his income tax returns were made for like periods as required by the revenue acts. We can no more segregate the varieties of the gambling operations for each year and make distinction thereon than we can say that, in a game of dealer’s choice poker, the winnings or losses of each hand dealt must be segregated because there was a variation in the form of game in each hand. In consequence of his gambling, Scaife [774]*774actually lost $25,205 in 1919 and $38,408 in 1920. For the reasons set forth in this opinion, these amounts can not be allowed as deductions.”
We believe that the Frey decision was unfortunate in that it disregarded two very real distinctions between the activities of the professional and the casual gambler. First, the casual gambler has no inevitable losses of the type included within the $8,800 figure in our illustrative case above, or representing the cost of goods raffled off in a lottery. Conceivably the casual gambler might never sustain a “loss” of any kind whatsoever.
Moreover the casual gambler ordinarily does not view his transactions as a yearly operation. That is to say, even if he should sustain losses he views them as isolated incidents unrelated to his winnings, not intending to exploit the laws of probability in such a way as to emerge with a favorable balance over an extended period of operations. Since the casual gambler views his bets individually as isolated incidents, there seems no good reason why the taxing authority should not do likewise. The professional, by way of contrast, founds his business directly upon the operation of the laws of probability over an extended period of time. When the taxing authority chooses to view his business as a series of unrelated individual transactions it does violence to the truth and arbitrarily deprives him of the benefit of the laws of probability upon which his enterprise is necessarily founded and without which it cannot succeed over any substantial period.
But regardless of the merits of the decision in the Frey case, a casual gambler was thereby permitted to exclude from gross income total winnings up to the amount of total losses in the same fashion as a professional gambler. In both McKenna and Frey, all of the gambling involved was illegal under state laws.
In addition, the professional gambler, if he suffered a net loss- over the year, could at that time deduct it under the provision equivalent to § 23(e) (1) of the Internal Revenue Code of 1939 where the losses were legally enforceable obligations.3
The casual gambler, however, had to bring himself within the provisions of § 23(e) (2) in order to get a deduction for net losses. Thus, in addition to showing that his gambling debts were legally enforceable, he also had to show that the transactions had been “entered into for profit.”
Losses sustained in illegal gambling operations, because not legally enforceable, were not allowed under either § 23(e) (1) or § 23(e) (2).
In 1934, with the law in this status, Congress enacted the provision that would become § 23(h) of the 1939 Code: “Losses from wagering transactions shall be allowed only to the extent of the gains from such transactions.” The effect of § 23(h) upon § 23(e) was determined in Humphrey v. Commissioner of Internal Revenue, 5 Cir., 1947, 162 F.2d 853, 175 A.L.R. 363, certiorari denied 1947, 332 U.S. 817, 68 S.Ct. 157, 92 L. Ed. 394. In that case the taxpayer was not a professional gambler. He engaged in only three wagering transactions. On two winning bets he made $2,140. On his’ losing bet he lost $3,000.
The Court held, in effect, that the $2,140 was gross income and that $2,140 of the $3,000 loss constituted an allowable deduction under § 23(h) despite the fact that the taxpayer made no showing that he could qualify under § 23(e) (2) as a “transaction entered into for prof[775]*775it.” The Court said at p. 855 of 162 F. 2d:
“ * * * We need not go behind Section 23 as it appears therein. It reads: ‘In computing net income there shall be allowed as deductions * * * (h) Wagering Losses. Losses from wagering transactions shall be allowed only to the extent of the gains from such transactions. * * * ’ There follow other classes of deductions through the letter (s). It is argued that Par. (h), Wagering Losses, is only a gloss on or an addition to Par. (e), and that (e) still governs such losses. We do not think so. If that had been the intent of Congress the matter of (h) would simply have been placed in (e). Instead the deductions (f) and (g) intervene, and are independent of (e); as is (h) which follows them. Each lettered paragraph authorizes a class of deductions. Wagering losses are made a class to themselves and ‘shall be allowed as deductions’, but ‘only to the extent of gains from such transactions.’ No longer need it be inquired whether the wagers were illegal so that the loss need not have been paid; nor what the state of mind of the taxpayer was as respects his purpose to win a profit. All wagers are by Par. (h) gathered into a class of their own and dealt with as the paragraph states. It is not necessary for the loser to prove he intended to win. The deduction for the losses ought to be allowed to the extent Sec. 23(h) requires.”
In other words, § 23(h) was not just an additional restriction upon § 23(e) but was in substitution therefor so far as wagering transactions are concerned.4
It is important to note, at this juncture, that § 23(h) had one important effect entirely separate and distinct from the one ordinarily assigned to it as exemplifying the congressional intent. It is clear that Congress intended to eliminate the difference between the treatment accorded to legal and illegal gambling losses. Legal net losses over the year, as we have seen, could be deducted under § 23(e). Illegal net losses could not be deducted at all. But regardless of legality or illegality, winnings were excluded from gross income up to the amount of total losses.
With the law in this condition, Congress could have eliminated the distinction between legal and illegal gambling losses simply by providing that no net gambling losses of any kind could be taken under § 23(e). This approach would have left the concept of gross income set forth in the McKenna and Frey cases unimpaired. Instead, it chose to employ the language of § 23(h). Accordingly, to the extent that § 23(h) controls, winning bets can no longer be excluded from gross income up to the amount of losing bets, since, if this were done, there could be no “gains” to which the “losses from wagering transactions” could be applied as a deduction. Therefore, it will be seen that, in addition to carrying out the legislative intent of eliminating the distinction between legal and illegal gambling losses, § 23(h) also had the effect of altering the then existing definition of gross income insofar as constitutional considerations would permit. That is, as to a casual gambler, the losses up to the amount of gambling winnings, formerly considered an exclusion from gross income under the Frey case, were now to be treated as a deduction.5 Further, the casual gambler’s use of losses was limited to that amount by the new subsection. But as to a pro[776]*776fessional, we believe that Congress lacked the power to undo the McKenna rule by statutory change and therefore that the new section could affect the professional only by barring his former right to deduct legal net gambling losses against other forms of income. Losses up to the amount of gambling winnings continue to be a reduction of the amount to be reported as gross income by the professional gambler. The holding of the Humphrey case, as distinguished from any dicta therein, is not inconsistent with this analysis, because taxpayer Humphrey was a casual gambler.
In Skeeles v. United States, 1951, 95 F.Supp. 242, 244, 118 Ct.Cl. 362, certio-rari denied 1951, 341 U.S. 948, 71 S.Ct. 1014, 95 L.Ed. 1371, which was an action for a refund of overpayments of income taxes, the Court said:
“This background is given as an aid to the interpretation of the statute. For the occasional wagerer no carry-over deduction is allowed. To sustain the contention of plaintiff it is necessary to extend to the professional gambler tax-deduction privileges that are not accorded to one who bets occasionally. To reach such a conclusion in the face of the public viewpoint and legislative purposes generally would require a clearly worded statute that precludes any other reasonable conclusion.”
The Court went on to decide that § 23(h) was applicable to the professional gambler, with the result that a net operating loss carry-over was denied. Accord, Of-futt v. C. I. R., 1951, 16 T.C. 1214.
We regard the decisions in the Skeeles and Offutt eases as not controlling. They dealt with attempts to utilize net gambling losses as an offset against the income of a different taxable year. As with the attempt to offset net gambling losses against other forms of income in the same taxable year, this question is properly a matter in which Congress could limit the professional as well as the casual gambler. On the other hand, the question in the present case is not one of defining deductions from gross income but rather of defining what is gross income. As already stated in this opinion, we believe there are inherent limitations upon the congressional power to deal with this matter.
To summarize briefly, we think that it is perhaps conceivable that Congress could, in its discretion, deny to a professional gambler the right to deduct from his total net winnings on winning races the total of his net losses on losing races in calculating the amount of his gross income. But we believe that Congress is without power to deny the professional gambler the right to offset his winnings on each race with his losses in that same race before coming to a “gain” of the type which constitutes gross income under § 22 of the Code. In other words, the appropriate unit for calculating his “gains” under § 22 may or may not be the net result over the yearly operation, but in any event it cannot be a unit which encompasses anything less than the total of his net profits (winning bets less losing bets) on every race.
We need not now decide which of these two theories constitutes the true definition of a professional gambler’s gross income. He has no gain at all within the meaning of § 22 unless his winnings in a particular race exceed his losses in that same race. Accordingly, it follows that the trial judge’s charge in regard to what constitutes the appellant’s gross income was erroneous.
It becomes unnecessary to consider the other contentions of the appellant in view of our ruling here.
The judgment of the district court is reversed.